Vision

Machine Vision
Toolbox
for MATLAB
Release 4
Peter Corke
2
Release 4.1
Release date July 2017
Licence LGPL
Toolbox home page http://www.petercorke.com/robot
Discussion group http://groups.google.com.au/group/robotics-tool-box
Copyright 2017
c Peter Corke
[email protected]
http://www.petercorke.com
Preface
This, the fourth major release of the Toolbox, repre-

senting nearly twenty years of continuous development.
This version corresponds to the second edition of the
book “Robotics, Vision & Control, second edition” pub-
lished in 2017.
The Machine Vision Toolbox (MVTB) provides many
functions that are useful in machine vision and vision-
based control. It is a somewhat eclectic collection
reflecting my personal interest in areas of photome-
try, photogrammetry, colorimetry. It includes over 100
functions spanning operations such as image file read-
ing and writing, acquisition, display, filtering, blob,
point and line feature extraction, mathematical mor-
phology, homographies, visual Jacobians, camera calibration and color space conver-
sion. The Toolbox, combined with MATLAB and a modern PC is a useful and conve-
nient environment for investigation of machine vision algorithms. For modest image
sizes the processing rate can be sufficiently “real-time” to allow for closed-loop control.
An image is usually treated as a rectangular array of pixel values – the natural datatype
for MATLAB– representing intensity or perhaps range. Many image operations such as
thresholding, filtering and statistics can be achieved with existing MATLAB functions.
The Toolbox extends this core functionality with M-files that implement functions and
classes, and mex-files for some compute intensive operations. This toolbox predates
all of the relevant Mathwork’s Toolboxes including Image Processing Toolbox R
(IPT)
R
and Computer Vision System Toolbox (CVST). MVTB is less complete than these
products but is open-source.
The code is written in a straightforward manner which allows for easy understanding,
perhaps at the expense of computational efficiency. If you feel strongly about computa-
tional efficiency then you can always rewrite the function to be more efficient, compile
the M-file using the MATLAB compiler, or create a MEX version.
The bulk of this manual is auto-generated from the comments in the MATLAB code
itself. For elaboration on the underlying principles, extensive illustrations and worked
examples please consult “Robotics, Vision & Control, second edition” which provides
a detailed discussion (720 pages, nearly 500 figures and over 1000 code examples) of
how to use the Toolbox functions to solve many types of problems in robotics.
Machine Vision Toolbox 4.1 for MATLAB3 Copyright Peter

c Corke 2017
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Functions by category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1 Introduction 6
1.1 Changes in MVTB 4 . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.1 Incompatible changes . . . . . . . . . . . . . . . . . . . . . . 6
1.1.2 New features . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.3 Enhancements . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 How to obtain the Toolbox . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 From .mltbx file . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.2 From .zip file . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.3 MATLAB OnlineTM . . . . . . . . . . . . . . . . . . . . . . 8
1.2.4 Simulink R
. . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.5 Documentation . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Compatible MATLAB versions . . . . . . . . . . . . . . . . . . . . . 9
1.4 Use in teaching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Use in research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7 Related software . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7.1 Image Processing Toolbox . . . . . . . . . . . . . . . . . . . 10
1.7.2 Computer Vision System Toolbox . . . . . . . . . . . . . . . 10
1.7.3 Octave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7.4 Robotics Toolbox . . . . . . . . . . . . . . . . . . . . . . . . 11
1.8 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Functions and classes 12

about . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
anaglyph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
apriltags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
AxisWebCamera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
BagOfWords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
blackbody . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
bresenham . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
camcald . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
CatadioptricCamera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
ccdresponse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
ccxyz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

c Corke 2017
CONTENTS CONTENTS
CentralCamera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
chi2inv_rtb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
cie_primaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
closest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
cmfrgb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
cmfxyz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
col2im . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
colnorm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
colordistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
colorize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
colorkmeans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
colorname . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
colorseg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
dtransform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
e2h . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
EarthView . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
edgelist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
epidist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
epiline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
FeatureMatch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
filt1d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
FishEyeCamera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
fmatrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
h2e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
hist2d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
hitormiss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
homline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
homography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
homtrans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
homwarp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Hough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
humoments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
ianimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
ibbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
iblobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
icanny . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
iclose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
icolor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
iconcat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
iconvolve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
icorner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
icp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
idecimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
idilate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
idisp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
idisplabel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
idouble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
iendpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
ierode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

c Corke 2017
CONTENTS CONTENTS
igamm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
igraphseg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
ihist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
iint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
iisum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
ilabel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
iline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
im2col . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
ImageSource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
imatch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
imeshgrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
imoments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
imono . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
imorph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
imser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
inormhist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
intgimage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
invcamcal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
iopen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
ipad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
ipaste . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
ipixswitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
iprofile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
ipyramid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
irank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
iread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
irectify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
ireplicate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
iroi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
irotate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
isamesize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
iscale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
iscalemax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
iscalespace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
iscolor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
ishomog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
ishomog2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
isift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
isimilarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
isize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
ismooth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
isobel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
isrot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
istereo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
istretch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
isurf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
isvec . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
ithin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
ithresh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
itrim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

c Corke 2017
CONTENTS CONTENTS
itriplepoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
ivar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
iwindow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
kcircle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
kdgauss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
kdog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
kgauss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
klaplace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
klog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
kmeans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
ksobel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
ktriangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
lambda2rg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
lambda2xy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
LineFeature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
loadspectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
luminos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
mkcube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
mkgrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
morphdemo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
Movie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
mpq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
mpq_poly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
ncc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
niblack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
npq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
npq_poly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
numcols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
numrows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
OrientedScalePointFeature . . . . . . . . . . . . . . . . . . . . . . . . . . 149
otsu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
peak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
peak2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
pickregion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
plot_arrow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
plot_box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
plot_circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
plot_ellipse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
plot_homline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
plot_point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
plot_poly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
plot_sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Plucker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
pnmfilt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
PointFeature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
polydiff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
radgrad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
ransac . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Ray3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
RegionFeature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

c Corke 2017
CONTENTS CONTENTS
rg_addticks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
rgb2xyz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
rluminos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
sad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
ScalePointFeature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
showcolorspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
showpixels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
SiftPointFeature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
SphericalCamera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
ssd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
stdisp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
SurfPointFeature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
tb_optparse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
testpattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Tracker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
tristim2cc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
upq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
upq_poly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
usefig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
VideoCamera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
VideoCamera_fg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
VideoCamera_IAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
xaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
xyzlabel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
yaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
YUV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
yuv2rgb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
yuv2rgb2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
zcross . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
zncc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
zsad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
zssd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

c Corke 2017
Functions by category
Color Image sources
blackbody . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Devices
ccdresponse . . . . . . . . . . . . . . . . . . . . . . . . 31
AxisWebCamera . . . . . . . . . . . . . . . . . . . . 14
ccxyz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
EarthView . . . . . . . . . . . . . . . . . . . . . . . . . . 52
cie_primaries . . . . . . . . . . . . . . . . . . . . . . . 44 ImageSource . . . . . . . . . . . . . . . . . . . . . . . 97
cmfrgb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Movie . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
cmfxyz . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 VideoCamera_IAT . . . . . . . . . . . . . . . . . 205
colordistance . . . . . . . . . . . . . . . . . . . . . . . 47 VideoCamera_fg . . . . . . . . . . . . . . . . . . . 203
colorname . . . . . . . . . . . . . . . . . . . . . . . . . . 49 VideoCamera . . . . . . . . . . . . . . . . . . . . . . 202
lambda2rg . . . . . . . . . . . . . . . . . . . . . . . . 136 YUV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
lambda2xy . . . . . . . . . . . . . . . . . . . . . . . . 137
loadspectrum . . . . . . . . . . . . . . . . . . . . . . 140 Test patterns
luminos . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
rg_addticks . . . . . . . . . . . . . . . . . . . . . . . . 181 mkcube . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
rgb2xyz . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 mkgrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
testpattern . . . . . . . . . . . . . . . . . . . . . . . . . 197
rluminos . . . . . . . . . . . . . . . . . . . . . . . . . . 181
showcolorspace . . . . . . . . . . . . . . . . . . . . 184
tristim2cc . . . . . . . . . . . . . . . . . . . . . . . . . 200 Monadic operators
yuv2rgb2 . . . . . . . . . . . . . . . . . . . . . . . . . 210
colorize . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
yuv2rgb . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
icolor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
igamm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
imono . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
inormhist . . . . . . . . . . . . . . . . . . . . . . . . . 104
istretch . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Camera models
Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Type changing

CatadioptricCamera . . . . . . . . . . . . . . . . . 28 idouble . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
CentralCamera . . . . . . . . . . . . . . . . . . . . . . 32 iint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
FishEyeCamera . . . . . . . . . . . . . . . . . . . . . 63
SphericalCamera. . . . . . . . . . . . . . . . . . .188
camcald . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Diadic operators
invcamcal . . . . . . . . . . . . . . . . . . . . . . . . . 105 ipixswitch . . . . . . . . . . . . . . . . . . . . . . . . . 107

c Corke 2017
CONTENTS CONTENTS
Spatial operators zncc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

zsad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Linear operators zssd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
icanny . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
iconvolve. . . . . . . . . . . . . . . . . . . . . . . . . . .81 Features
ismooth . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
isobel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 Region features
radgrad . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
RegionFeature . . . . . . . . . . . . . . . . . . . . . 176
colorkmeans . . . . . . . . . . . . . . . . . . . . . . . . 48
Kernels colorseg . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
ibbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
kcircle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 iblobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
kdgauss . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 igraphseg . . . . . . . . . . . . . . . . . . . . . . . . . . 92
kdog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 ilabel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
kgauss . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 imoments . . . . . . . . . . . . . . . . . . . . . . . . . 100
klaplace . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 imser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
klog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 ithresh . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
ksobel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 niblack . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
ktriangle . . . . . . . . . . . . . . . . . . . . . . . . . . 136 otsu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Non-linear operators Line features

dtransform . . . . . . . . . . . . . . . . . . . . . . . . . 51 Hough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
irank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 LineFeature . . . . . . . . . . . . . . . . . . . . . . . 137
ivar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
iwindow . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Point features
Morphological FeatureMatch . . . . . . . . . . . . . . . . . . . . . . . 57
OrientedScalePointFeature . . . . . . . . . . 149
hitormiss . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 PointFeature . . . . . . . . . . . . . . . . . . . . . . . 168
iclose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 ScalePointFeature . . . . . . . . . . . . . . . . . . 182
idilate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 SiftPointFeature . . . . . . . . . . . . . . . . . . . 185
iendpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 SurfPointFeature . . . . . . . . . . . . . . . . . . . 193
ierode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 icorner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
imorph . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 iscalemax . . . . . . . . . . . . . . . . . . . . . . . . . 115
iopen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 iscalespace . . . . . . . . . . . . . . . . . . . . . . . . 115
ithin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 isift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
itriplepoint . . . . . . . . . . . . . . . . . . . . . . . . 129 isurf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
morphdemo . . . . . . . . . . . . . . . . . . . . . . . 142
Other features
Similarity
apriltags . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
imatch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 hist2d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
isimilarity . . . . . . . . . . . . . . . . . . . . . . . . . 119 ihist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
ncc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 iprofile . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
sad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 peak2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
ssd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 peak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

c Corke 2017
CONTENTS CONTENTS
Multiview Image generation

iconcat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Geometric iline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
ipaste . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
epidist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
epiline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
fmatrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
homography . . . . . . . . . . . . . . . . . . . . . . . . 69 Moments
humoments . . . . . . . . . . . . . . . . . . . . . . . . . 74
mpq_poly . . . . . . . . . . . . . . . . . . . . . . . . . 145
Stereo mpq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
npq_poly . . . . . . . . . . . . . . . . . . . . . . . . . . 147
anaglyph . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
npq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
irectify . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
upq_poly . . . . . . . . . . . . . . . . . . . . . . . . . . 201
istereo . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
upq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
stdisp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Plotting
Image sequence
plot_arrow . . . . . . . . . . . . . . . . . . . . . . . . 154
BagOfWords . . . . . . . . . . . . . . . . . . . . . . . 16 plot_box . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Tracker . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 plot_circle . . . . . . . . . . . . . . . . . . . . . . . . 155
ianimate . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 plot_ellipse . . . . . . . . . . . . . . . . . . . . . . . . 156
plot_homline . . . . . . . . . . . . . . . . . . . . . . 157
plot_point . . . . . . . . . . . . . . . . . . . . . . . . . 158
Shape changing plot_poly . . . . . . . . . . . . . . . . . . . . . . . . . 159
plot_sphere . . . . . . . . . . . . . . . . . . . . . . . . 160
homwarp . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
idecimate . . . . . . . . . . . . . . . . . . . . . . . . . . 85
ipad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
ipyramid . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Homogeneous coordinates
ireplicate . . . . . . . . . . . . . . . . . . . . . . . . . . 112 e2h . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
iroi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 h2e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
irotate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 homline . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
isamesize . . . . . . . . . . . . . . . . . . . . . . . . . 114 homtrans . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
iscale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
itrim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Homogeneous coordinates in
2D
Utility
ishomog2 . . . . . . . . . . . . . . . . . . . . . . . . . 117
Image utility
idisplabel . . . . . . . . . . . . . . . . . . . . . . . . . . 88 Homogeneous coordinates in
idisp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 3D
iread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
pnmfilt . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 ishomog . . . . . . . . . . . . . . . . . . . . . . . . . . 116
showpixels . . . . . . . . . . . . . . . . . . . . . . . . 185 isrot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

c Corke 2017
CONTENTS CONTENTS
3D geometry col2im . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
colnorm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Plucker . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Ray3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
filt1d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
icp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .84
im2col . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
imeshgrid . . . . . . . . . . . . . . . . . . . . . . . . . 100
Integral image iscolor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
isize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
iisum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 isvec . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
intgimage . . . . . . . . . . . . . . . . . . . . . . . . . 104 kmeans . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
numcols . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Edges and lines numrows . . . . . . . . . . . . . . . . . . . . . . . . . . 148
pickregion . . . . . . . . . . . . . . . . . . . . . . . . 153
bresenham . . . . . . . . . . . . . . . . . . . . . . . . . 20 polydiff . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
edgelist . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 ransac . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
tb_optparse . . . . . . . . . . . . . . . . . . . . . . . . 196
usefig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
General xaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
about . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 xyzlabel . . . . . . . . . . . . . . . . . . . . . . . . . . 207
chi2inv_rtb . . . . . . . . . . . . . . . . . . . . . . . . . 43 yaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
closest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 zcross . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

c Corke 2017
Chapter 1
Introduction
1.1 Changes in MVTB 4
MVTB 4 is largely backward compatible with MVTB 3.
1.1.1 Incompatible changes
• xycolorspace replaced with showcolorspace(’xy’). It can also dis-

play Lab colorspace.
• iconv changed to iconvolve
• solar.dat has the units changed from W/m2 /nm to W/m2 /m.
• RegionFeature property shape has been renamed to aspect
• Data files that were previously in the folder private but are now in data.
• Options of the form ’Tcam’, ’Tobj’, ’T0’ or ’Tf’ are now respectively
’pose’, ’objpose’, ’pose0’ or ’posef’.
1.1.2 New features
• For a vector of RegionFeature objects all the properties can now be extracted
as vectors.
• Bundle adjustment is supported by the new BundleAdjust class.
• The folder symbolic contains Live Scripts that demonstrate use of the MAT-
LAB Symbolic Math ToolboxTM for deriving Jacobians related to bundle adjust-
ment, image Jacobian for visual servoing and Gaussian kernels.

c Corke 2017
1.2. HOW TO OBTAIN THE TOOLBOX CHAPTER 1. INTRODUCTION
1.1.3 Enhancements
• cmf2xyz for spectral input multiplies by ∆λ

• color data
1.2 How to obtain the Toolbox
The Machine Vision Toolbox is freely available from the Toolbox home page at
http://www.petercorke.com
The file is available in MATLABtoolbox format (.mltbx) or zip format (.zip).
1.2.1 From .mltbx file
Since MATLAB R2014b toolboxes can be packaged as, and installed from, files with
the extension .mltbx. Download the most recent version of robot.mltbx or
vision.mltbx to your computer. Using MATLAB navigate to the folder where
you downloaded the file and double-click it (or right-click then select Install). The
Toolbox will be installed within the local MATLAB file structure, and the paths will be
appropriately configured for this, and future MATLAB sessions.
1.2.2 From .zip file
Download the most recent version of robot.zip or vision.zip to your computer. Use
your favourite unarchiving tool to unzip the files that you downloaded. To add the
Toolboxes to your MATLAB path execute the command
>> addpath RVCDIR ;
>> startup_rvc
where RVCDIR is the full pathname of the folder where the folder rvctools was
created when you unzipped the Toolbox files. The script startup_rvc adds various
subfolders to your path and displays the version of the Toolboxes. After installation
the files for both Toolboxes reside in a top-level folder called rvctools and beneath
this are a number of folders:
robot The Robotics Toolbox
vision The Machine Vision Toolbox
common Utility functions common to the Robotics and Machine Vision Toolboxes
simulink Simulink blocks for robotics and vision, as well as examples
contrib Code written by third-parties
A menu-driven demonstration can be invoked by
>> mvtbdemo
The MVTB distribution includes the code and example images necessary to do almost
all the examples in the Robotics, Vision & Control book. Additional files are available:

c Corke 2017
CHAPTER 1. INTRODUCTION 1.2. HOW TO OBTAIN THE TOOLBOX
• contrib.zip A small number of Toolbox functions depend on third party

code which is included in this file. Please note and respect the licence conditions
associated with these packages. Those functions are: igraphseg, imser, and
CentralCamera.estpose.
• contrib2.zip Additional third party code for the functions: isift, and
isurf. Note that the code here is slightly modified version of the open-source
packages.
• images2.zip This is a large file (150MB) containing the mosaic, campus,

bridge-l and campus sequences which support the examples in Sections 14.6,
14.7 and 14.8 respectively.
If you already have the Robotics Toolbox installed then download the zip file(s) to the
directory above the existing rvctools directory and then unzip them. The files from
these zip archives will properly interleave with the Robotics Toolbox files.
Ensure that the folder rvctools is on your MATLAB search path. You can do this
by issuing the addpath command at the MATLAB prompt. Then issue the com-
mand startup_rvc and it will add a number of paths to your MATLAB search
path. You need to setup the path every time you start MATLAB but you can automate
this by setting up environment variables, editing your startup.m script by pressing
the “Update Toolbox Path Cache" button under MATLAB General preferences.
1.2.3 MATLAB OnlineTM
The Toolbox works well with MATLAB OnlineTM which lets you access a MATLAB
session from a web browser, tablet or even a phone. The key is to get the MVTB
files into the filesystem associated with your Online account. The easiest way to do
this is to install MATLAB DriveTM from MATLAB File Exchange or using the Get
Add-Ons option from the MATLAB GUI. This functions just like Google Drive or
Dropbox, a local filesystem on your computer is synchronized with your MATLAB
Online account. Copy the MVTB files into the local MATLAB Drive cache and they
will soon be synchronized, invoke startup_rvc to setup the paths and you are ready
to machine vision on your mobile device or in a web browser.
MATLAB Online does not support MEX files.
1.2.4 Simulink
R
Simulink R
is the block diagram simulation environment for MATLAB. The following
Simulink models are included with the Toolbox, but rely on having RTB installed.

c Corke 2017
1.3. COMPATIBLE MATLAB VERSIONS CHAPTER 1. INTRODUCTION
General
sl_ibvs Classical IBVS
sl_partitioned XY/Z partioned IBVS
Robot manipulator arm
sl_arm_ibvs Servo a 6DOF robot arm
Mobile ground robot
sl_drivepose_vs Drive nonholonomic robot to a pose
sl_mobile_vs Drive a holonomic vehicle to a pose
Flying robot
sl_quadrotor_vs Control visual servoing to a target
1.2.5 Documentation
This document vision.pdf is a comprehensive manual that describes all functions

in the Toolbox. It is auto-generated from the comments in the MATLAB code and is
fully hyperlinked: to external web sites, the table of content to functions, and the “See
also” functions to each other.
The same documentation is available online in alphabetical order at http://www.
petercorke.com/MVTB/r4/html/index_alpha.html or by category at http:
//www.petercorke.com/MVTB/r4/html/index.html. Documentation is
also available via the MATLAB help browser, under supplemental software, as “Ma-
chine Vision Toolbox".
1.3 Compatible MATLAB versions
The Toolbox has been tested under R2016b and R2017aPRE. Compatibility problems
are increasingly likely the older your version of MATLAB is.
1.4 Use in teaching
This is definitely encouraged! You are free to put the PDF manual (robot.pdf or
the web-based documentation html/*.html on a server for class use. If you plan to
distribute paper copies of the PDF manual then every copy must include the first two
pages (cover and licence).
Link to other resources such as MOOCs or the Robot Academy can be found at www.
petercorke.com/moocs.
1.5 Use in research
If the Toolbox helps you in your endeavours then I’d appreciate you citing the Toolbox
when you publish. The details are:

c Corke 2017
CHAPTER 1. INTRODUCTION 1.6. SUPPORT
@book{Corke17a,
Author = {Peter I. Corke},
Note = {ISBN 978-3-319-54413-7},
Edition = {Second},
Publisher = {Springer},
Title = {Robotics, Vision \& Control: Fundamental Algorithms in {MATLAB}},
Year = {2017}}
or
P.I. Corke, Robotics, Vision & Control: Fundamental Algorithms in MAT-
LAB. Second edition. Springer, 2017. ISBN 978-3-319-54413-7.
which is also given in electronic form in the CITATION file.
1.6 Support
There is no support! This software is made freely available in the hope that you find it
useful in solving whatever problems you have to hand. I am happy to correspond with
people who have found genuine bugs or deficiencies but my response time can be long
and I can’t guarantee that I respond to your email.
I can guarantee that I will not respond to any requests for help with assignments
or homework, no matter how urgent or important they might be to you. That’s
what your teachers, tutors, lecturers and professors are paid to do.
You might instead like to communicate with other users via the Google Group called
“Robotics and Machine Vision Toolbox”
http://tiny.cc/rvcforum
which is a forum for discussion. You need to signup in order to post, and the signup
process is moderated by me so allow a few days for this to happen. I need you to write a
few words about why you want to join the list so I can distinguish you from a spammer
or a web-bot.
1.7 Related software
1.7.1 Image Processing Toolbox
The Image Processing ToolboxTM (IPT) from MathWorks is an official and supported
product. This toolbox includes a comprehensive set of image processing operations.
1.7.2 Computer Vision System Toolbox
The Computer Vision System ToolboxTM (CVST) from MathWorks is an official and
supported product. System toolboxes (see also the Computer Vision System Toolbox)
are aimed at developers of systems. This toolbox includes a comprehensive set of

c Corke 2017
1.8. ACKNOWLEDGEMENTS CHAPTER 1. INTRODUCTION
feature detectors and descriptors, and be used with Simulink to conveniently generate
machine vision pipelines that can run in target hardware.
1.7.3 Octave
GNU Octave (www.octave.org) is an impressive piece of free software that implements

a language that is close to, but not the same as, MATLAB.
1.7.4 Robotics Toolbox
Robotics toolbox (RTB) for MATLAB provides a very wide range of useful robotics
functions and is used to illustrate principals in the Robotics, Vision & Control book.
You can obtain this from http://www.petercorke.com/robot.
1.8 Acknowledgements
This release includes functions for computing image plane homographies and the fun-
damental matrix, contributed by Nuno Alexandre Cid Martins of I.S.R., Coimbra.
RANSAC code by Peter Kovesi; pose estimation by Francesco Moreno-Noguer, Vin-
cent Lepetit, Pascal Fua at the CVLab-EPFL; color space conversions by Pascal Ge-
treuer; numerical routines for geometric vision by various members of the Visual Ge-
ometry Group at Oxford (from the web site of the Hartley and Zisserman book1 ; the k-
means and MSER algorithms by Andrea Vedaldi and Brian Fulkerson;the graph-based
image segmentation software by Pedro Felzenszwalb; and the SURF feature detec-
tor by Dirk-Jan Kroon at U. Twente. The Camera Calibration Toolbox by Jean-Yves
Bouguet is used unmodified.Functions such as SURF, MSER, graph-based segmenta-
tion and pose estimation are based on great code Some of the MEX file use some really
neat macros that were part of the package VISTA Copyright 1993, 1994 University of
British Columbia. See the file CONTRIB for details.
1 http://www.robots.ox.ac.uk/~vgg/hzbook

c Corke 2017
Chapter 2
Functions and classes
about
Compact display of variable type
about(x) displays a compact line that describes the class and dimensions of x.
about x as above but this is the command rather than functional form
Examples
>> a=1;
>> about a
a [double] : 1x1 (8 bytes)
>> a = rand(5,7);
>> about a
a [double] : 5x7 (280 bytes)
See also
whos
anaglyph
Convert stereo images to an anaglyph image
a = anaglyph(left, right) is an anaglyph image where the two images of a stereo pair
are combined into a single image by coding them in two different colors. By default

c Corke 2017
CHAPTER 2. FUNCTIONS AND CLASSES
the left image is red, and the right image is cyan.

anaglyph(left, right) as above but display the anaglyph.
a = anaglyph(left, right, color) as above but the string color describes the color coding
as a string with 2 letters, the first for left, the second for right, and each is one of:
‘r’ red
‘g’ green
‘b’ green
‘c’ cyan
‘m’ magenta
a = anaglyph(left, right, color, disp) as above but allows for disparity correction. If
disp is positive the disparity is increased, if negative it is reduced. These adjustments
are achieved by trimming the images. Use this option to make the images more nat-
ural/comfortable to view, useful if the images were captured with a stereo baseline
significantly different the human eye separation (typically 65mm).
Example
Load the left and right images

L = iread(’rocks2-l.png’, ’reduce’, 2);
R = iread(’rocks2-r.png’, ’reduce’, 2);
then display the anaglyph for viewing with red-cyan glasses

anaglyph(L, R);
References
• Robotics, Vision & Control, Section 14.3, P. Corke, Springer 2011.
See also
stdisp
apriltags
Read April tags from image
tags = apriltags(im) is a vector of structures that describe each of the April tags found
within the image IM.

c Corke 2017
Elements of the structure are:
.id decoded id of the tag in the range 1-255

.hamming number of corrected id bits, 0 is best
.goodness tag constrast quality
.margin decision margin, high is better
.H homography matrix (3 × 3) describing the projection from an “ideal” tag (with corners
at (-1,-1), (1,-1), (1,1), and (-1,1)) to pixels in the image
.centre centre of the tag in the image (2 × 1)
.corners corners of the tag in the image (2 × 4)
Notes
• implementation is a mex file

• the options refine_decode and refine_pose are both enabled.
• the image must be uint8 or double (grey level range 0 to 1).
• only tag family tag36h11 is supported.
Author
• apriltags is open-source software from University of Michigan

• details at https://april.eecs.umich.edu/software/apriltag.html
• This wrapper by Peter Corke
AxisWebCamera
Image from Axis webcam
A concrete subclass of ImageSource that acquires images from a web camera built by
Axis Communications (www.axis.com).
Methods
grab Aquire and return the next image

size Size of image
close Close the image source
char Convert the object parameters to human readable string

c Corke 2017
See also
ImageSource, Video
AxisWebCamera.AxisWebCamera
Axis web camera constructor
a = AxisWebCamera(url, options) is an AxisWebCamera object that acquires im-

ages from an Axis Communications (www.axis.com) web camera.
Options
‘uint8’ Return image with uint8 pixels (default)

‘float’ Return image with float pixels
‘double’ Return image with double precision pixels
‘grey’ Return greyscale image
‘gamma’, G Apply gamma correction with gamma=G
‘scale’, S Subsample the image by S in both directions.
‘resolution’, S Obtain an image of size S=[W H].
Notes:
• The specified ‘resolution’ must match one that the camera is capable of, other-
wise the result is not predictable.
AxisWebCamera.char
Convert to string
A.char() is a string representing the state of the camera object in human readable form.
See also
AxisWebCamera.display

c Corke 2017
AxisWebCamera.close
Close the image source
A.close() closes the connection to the web camera.
AxisWebCamera.grab
Acquire image from the camera
im = A.grab() is an image acquired from the web camera.
Notes
• Some web cameras have a fixed picture taking interval, and this function will
return the most recently captured image held in the camera.
BagOfWords
Bag of words class
The BagOfWords class holds sets of features for a number of images and supports
image retrieval by comparing new images with those in the ‘bag’.
Methods
isword Return all features assigned to word

occurrences Return number of occurrences of word
remove_stop Remove stop words
wordvector Return word frequency vector
wordfreq Return words and their frequencies
similarity Compare two word bags
contains List the images that contain a word
exemplars Display examples of word support regions
display Display the parameters of the bag of words
char Convert the parameters of the bag of words to a string
Properties

c Corke 2017
K The number of clusters specified

nstop The number of stop words specified
nimages The number of images in the bag
Reference
J.Sivic and A.Zisserman, “Video Google: a text retrieval approach to object matching
in videos”, in Proc. Ninth IEEE Int. Conf. on Computer Vision, pp.1470-1477, Oct.
2003.
See also
PointFeature
BagOfWords.BagOfWords
Create a BagOfWords object
b = BagOfWords(f, k) is a new bag of words created from the feature vector f and with
k words. f can also be a cell array, as produced by ISURF() for an image sequence.
The features are sorted into k clusters and each cluster is termed a visual word.
b = BagOfWords(f, b2) is a new bag of words created from the feature vector f but
clustered to the words (and stop words) from the existing bag b2.
Notes
• Uses the MEX function vl_kmeans to perform clustering (vlfeat.org).
See also
PointFeature, isurf
BagOfWords.char
Convert to string
s = B.char() is a compact string representation of a bag of words.

c Corke 2017
BagOfWords.contains
Find images containing word
k = B.contains(w) is a vector of the indices of images in the sequence that contain one
or more instances of the word w.
BagOfWords.display
Display value
B.display() displays the parameters of the bag in a compact human readable form.
Notes
• This method is invoked implicitly at the command line when the result of an
expression is a BagOfWords object and the command has no trailing semicolon.
See also
BagOfWords.char
BagOfWords.exemplars
display exemplars of words
B.exemplars(w, images, options) displays examples of the support regions of the

words specified by the vector w. The examples are displayed as a table of thumb-
nail images. The original sequence of images from which the features were extracted
must be provided as images.
Options
‘columns’, N Number of columns to display (default 10)

‘maxperimage’, M Maximum number of exemplars to display from any one image (default 2)
‘width’, w Width of each thumbnail [pixels] (default 50)

c Corke 2017
BagOfWords.isword
Features from words
f = B.isword(w) is a vector of feature objects that are assigned to any of the word w. If
w is a vector of words the result is a vector of features assigned to all the words in w.
BagOfWords.occurrence
Word occurrence
n = B.occurrence(w) is the number of occurrences of the word w across all features in

the bag.
BagOfWords.remove_stop
Remove stop words
B.remove_stop(n) removes the n most frequent words (the stop words) from the bag.
All remaining words are renumbered so that the word labels are consecutive.
BagOfWords.wordfreq
Word frequency statistics
[w,n] = B.wordfreq() is a vector of word labels w and the corresponding elements of

n are the number of occurrences of that word.
BagOfWords.wordvector
Word frequency vector
wf = B.wordvector(J) is the word frequency vector for the Jth image in the bag. The
vector is K × 1 and the angle between any two WFVs is an indication of image simi-
larity.

c Corke 2017
Notes
• The word vector is expensive to compute so a lazy evaluation is performed on

the first call to this function
blackbody
Compute blackbody emission spectrum
E = blackbody(lambda, T) is the blackbody radiation power density [W/m3 ] at the

wavelength lambda [m] and temperature T [K].
If lambda is a column vector (N × 1), then E is a column vector (N × 1) of blackbody
radiation power density at the corresponding elements of lambda.
Example
l = [380:10:700]’*1e-9; % visible spectrum
e = blackbody(l, 6500); % emission of sun
plot(l, e)
References
bresenham
Generate a line
p = bresenham(x1, y1, x2, y2) is a list of integer coordinates (2 × N) for points lying
on the line segment joining the integer coordinates (x1,y1) and (x2,y2).
p = bresenham(p1, p2) as above but p1=[x1; y1] and p2=[x2; y2].
Notes
• Endpoint coordinates must be integer values.

c Corke 2017
Author
• Based on code by Aaron Wetzler
See also
icanvas
camcald
Camera calibration from data points
C = camcald(d) is the camera matrix (3 × 4) determined by least squares from corre-

sponding world and image-plane points. d is a table of points with rows of the form
[X Y Z U V] where (X,Y,Z) is the coordinate of a world point and [U,V] is the corre-
sponding image plane coordinate.
[C,E] = camcald(d) as above but E is the maximum residual error after back substitu-
tion [pixels].
Notes:
• This method assumes no lense distortion affecting the image plane coordinates.
See also
CentralCamera
Camera
Camera superclass
An abstract superclass for Toolbox camera classes.
Methods
plot plot projection of world point to image plane

hold control figure hold for image plane window
ishold test figure hold for image plane

c Corke 2017
clf clear image plane

figure figure holding the image plane
mesh draw shape represented as a mesh
point draw homogeneous points on image plane
homline draw homogeneous lines on image plane
lineseg draw line segment defined by points
plot_camera draw camera in world view
rpy set camera attitude
move clone Camera after motion
centre get world coordinate of camera centre
delete object destructor
char convert camera parameters to string
display display camera parameters
Properties (read/write)
npix image dimensions (2 × 1)

pp principal point (2 × 1)
rho pixel dimensions (2 × 1) in metres
T camera pose as homogeneous transformation
Properties (read only)
nu number of pixels in u-direction

nv number of pixels in v-direction
u0 principal point u-coordinate
v0 principal point v-coordinate
Notes
• Camera is a reference object.

• Camera objects can be used in vectors and arrays
• This is an abstract class and must be subclassed and a project() method defined.
• The object can create a window to display the Camera image plane, this window
is protected and can only be accessed by the plot methods of this object.
• The project method is implemented by the concrete subclass.
See also
CentralCamera, SphericalCamera, FishEyeCamera, CatadiptricCamera

c Corke 2017
Camera.Camera
Create camera object
Constructor for abstact Camera class, used by all subclasses.

C = Camera(options) creates a default (abstract) camera with null parameters.
Options
‘name’, N Name of camera

‘image’, IM Load image IM to image plane
‘resolution’, N Image plane resolution: N × N or N=[W H]
‘sensor’, S Image sensor size in metres (2 × 1) [metres]
‘centre’, P Principal point (2 × 1)
‘pixel’, S Pixel size: S × S or S=[W H]
‘noise’, SIGMA Standard deviation of additive Gaussian noise added to returned image projections
‘pose’, T Pose of the camera as a homogeneous transformation
‘color’, C Color of image plane background (default [1 1 0.8])
Notes
• Normally the class plots points and lines into a set of axes that represent the
image plane. The ‘image’ option paints the specified image onto the image plane
and allows points and lines to be overlaid.
See also
CentralCamera, FisheyeCamera, CatadioptricCamera, SphericalCamera
Camera.centre
Get camera position
p = C.centre() is the 3-dimensional position of the camera centre (3 × 1).
Camera.char
Convert to string
s = C.char() is a compact string representation of the camera parameters.

c Corke 2017
Camera.clf
Clear the image plane
C.clf() removes all graphics from the camera’s image plane.
Camera.delete
Camera object destructor
C.delete() destroys all figures associated with the Camera object and removes the
object.
disp(’delete camera object’);
Camera.display
Display value
C.display() displays a compact human-readable representation of the camera parame-

ters.
Notes
expression is a Camera object and the command has no trailing semicolon.
See also
Camera.char
Camera.figure
Return figure handle
H = C.figure() is the handle of the figure that contains the camera’s image plane graph-
ics.

c Corke 2017
Camera.hold
Control hold on image plane graphics
C.hold() sets “hold on” for the camera’s image plane.

C.hold(H) hold mode is set on if H is true (or > 0), and off if H is false (or 0).
Camera.homline
Plot homogeneous lines on image plane
C.homline(L) plots lines on the camera image plane which are defined by columns of
L (3 × N) considered as lines in homogeneous form: a.u + b.v + c = 0.
Camera.ishold
Return image plane hold status
H = C.ishold() returns true (1) if the camera’s image plane is in hold mode, otherwise
false (0).
Camera.lineseg
handle for this camera image plane
Camera.mesh
Plot mesh object on image plane
C.mesh(x, y, z, options) projects a 3D shape defined by the matrices x, y, z to the image

plane and plots them. The matrices x, y, z are of the same size and the corresponding
elements of the matrices define 3D points.
Options
‘objpose’, T Transform all points by the homogeneous transformation T before projecting them to
the camera image plane.

c Corke 2017
‘pose’, T Set the camera pose to the homogeneous transformation T before projecting points to
the camera image plane. Temporarily overrides the current camera pose C.T.
Additional arguments are passed to plot as line style parameters.
See also
mesh, cylinder, sphere, mkcube, Camera.plot, Camera.hold, Camera.clf
Camera.move
Instantiate displaced camera
C2 = C.move(T) is a new camera object that is a clone of C but its pose is displaced
by the homogeneous transformation T with respect to the current pose of C.
Camera.plot
Plot points on image plane
C.plot(p, options) projects world points p (3 × N) to the image plane and plots them. If
p is 2×N the points are assumed to be image plane coordinates and are plotted directly.
uv = C.plot(p) as above but returns the image plane coordinates uv (2 × N).
• If p has 3 dimensions (3 × N × S) then it is considered a sequence of point sets
and is displayed as an animation.
C.plot(L, options) projects the world lines represented by the array of Plucker objects
(1 × N) to the image plane and plots them.
li = C.plot(L, options) as above but returns an array (3 × N) of image plane lines in
homogeneous form.
Options
‘objpose’, T Transform all points by the homogeneous transformation T before
projecting them to the camera image plane.
the camera image plane. Overrides the current camera pose C.T.
‘fps’, N Number of frames per second for point sequence display

c Corke 2017
‘sequence’ Annotate the points with their index

‘textcolor’, C Text color for annotation (default black)
‘textsize’, S Text size for annotation (default 12)
‘drawnow’ Execute MATLAB drawnow function
Additional options are considered MATLAB linestyle parameters and are passed di-
rectly to plot.
See also
Camera.mesh, Camera.hold, Camera.clf, Plucker
Camera.plot_camera
Display camera icon in world view
C.plot_camera(options) draw a camera as a simple 3D model in the current figure.
Options
‘pose’, T Camera displayed in pose T (homogeneous transformation 4 × 4)

‘scale’, S Overall scale factor (default 0.2 x maximum axis dimension)
‘color’, C Camera body color (default blue)
‘frustum’ Draw the camera as a frustrum (pyramid mesh)
‘solid’ Draw a non-frustrum camera as a solid (default)
‘mesh’ Draw a non-frustrum camera as a mesh
‘label’ Show the camera’s name next to the camera
Notes
• The graphic handles are stored within the Camera object.

• A line between the red faces is parallel to the x-axis, between the green faces is
parallel to the y-axis.
Camera.point
Plot homogeneous points on image plane
C.point(p) plots points on the camera image plane which are defined by columns of p
(3 × N) considered as points in homogeneous form.

c Corke 2017
Camera.rpy
Set camera attitude
C.rpy(R, p, y) sets the camera attitude to the specified roll-pitch-yaw angles.

C.rpy(rpy) as above but rpy=[R,p,y].
CatadioptricCamera
Catadioptric camera class
A concrete class for a catadioptric camera, subclass of Camera.
Methods
project project world points to image plane
plot plot/return world point on image plane

hold control hold for image plane
line draw homogeneous lines on image plane
plot_camera draw camera

move copy of Camera after motion


c Corke 2017
npix image dimensions in pixels (2 × 1)

pp intrinsic: principal point (2 × 1)
rho intrinsic: pixel dimensions (2 × 1) [metres]
f intrinsic: focal length [metres]
p intrinsic: tangential distortion parameters
T extrinsic: camera pose as homogeneous transformation

Notes

See also
CentralCamera, Camera
CatadioptricCamera.CatadioptricCamera
Create central projection camera object
C = CatadioptricCamera() creates a central projection camera with canonic parame-

ters: f=1 and name=’canonic’.
C = CatadioptricCamera(options) as above but with specified parameters.
Options

‘focal’, F Focal length (metres)
‘default’ Default camera parameters: 1024 × 1024, f=8mm, 10um pixels, camera at origin,
optical axis is z-axis, u- and v-axes parallel to x- and y-axes respectively.
‘projection’, M Catadioptric model: ‘equiangular’ (default), ‘sine’, ‘equisolid’, ‘stereographic’
‘k’, K Parameter for the projection model

c Corke 2017
‘maxangle’, A The maximum viewing angle above the horizontal plane.

‘resolution’, N Image plane resolution: N × N or N=[W H].
‘sensor’, S Image sensor size in metres (2 × 1)
‘pixel’, S Pixel size: S × S or S=[W H].
Notes
• The elevation angle range is from -pi/2 (below the mirror) to maxangle above the
horizontal plane.
See also
Camera, FisheyeCamera, CatadioptricCamera, SphericalCamera
CatadioptricCamera.project
Project world points to image plane
uv = C.project(p, options) are the image plane coordinates for the world points p.
The columns of p (3 × N) are the world points and the columns of uv (2 × N) are the
corresponding image plane points.
Options
‘pose’, T Set the camera pose to the pose T (homogeneous transformation (4×4) or SE3) before
projecting points to the camera image plane. Temporarily overrides the current camera
pose C.T.
‘objpose’, T Transform all points by the pose T (homogeneous transformation (4 × 4) or SE3)
before projecting them to the camera image plane.
See also
FishEyeCamera.plot, Camera.plot

c Corke 2017
ccdresponse
CCD spectral response
R = ccdresponse(lambda) is the spectral response of a typical silicon imaging sen-

sor at the wavelength lambda [m]. The response is normalized in the range 0 to 1.
If lambda is a vector then R is a vector of the same length whose elements are the
response at the corresponding element of lambda.
Notes
• Deprecated, use loadspectrum(lambda, ‘ccd’) instead.
References
• An ancient Fairchild data book for a silicon sensor.

See also
rluminos
ccxyz
XYZ chromaticity coordinates
xyz = ccxyz(lambda) is the xyz-chromaticity coordinates (3 × 1) for illumination at

wavelength lambda. If lambda is a vector (N × 1) then each row of xyz (N × 3) is the
xyz-chromaticity of the corresponding element of lambda.
xyz = ccxyz(lambda, E) is the xyz-chromaticity coordinates (N ×3) for an illumination
spectrum E (N × 1) defined at corresponding wavelengths lambda (N × 1).
References

c Corke 2017
See also
cmfxyz
CentralCamera
Perspective camera class
A concrete class for a central-projection perspective camera, a subclass of Camera.

The camera coordinate system is:
0------------> u X
|
|
| + (principal point)
|
| Z-axis is into the page.
v Y
This camera model assumes central projection, that is, the focal point is at z=0 and the
image plane is at z=f. The image is not inverted.
Methods
project project world points and lines

K camera intrinsic matrix
C camera matrix
H camera motion to homography
invH decompose homography
F camera motion to fundamental matrix
E camera motion to essential matrix
invE decompose essential matrix
fov field of view
ray Ray3D corresponding to point
centre projective centre
normalized convert image plane coordinate to normalized coordinates
plot plot projection of world point on image plane
plot_camera draw camera in world view
plot_line_tr draw line in theta/rho format

c Corke 2017
plot_epiline draw epipolar line

flowfield compute optical flow
visjac_p image Jacobian for point features
visjac_p_polar image Jacobian for point features in polar coordinates
visjac_l image Jacobian for line features
visjac_e image Jacobian for ellipse features
derivs point and camera motion Jacobians for bundle adjustment
move clone Camera after motion
estpose estimate pose

rho intrinsic: pixel dimensions (2 × 1) in metres
f intrinsic: focal length
k intrinsic: radial distortion vector
p intrinsic: tangential distortion parameters
distortion intrinsic: camera distortion [k1 k2 k3 p1 p2]

Notes
See also
Camera

c Corke 2017
CentralCamera.CentralCamera
Create central projection camera object
C = CentralCamera() creates a central projection camera with canonic parameters:

f=1 and name=’canonic’.
C = CentralCamera(options) as above but with specified parameters.
Options

‘focal’, F Focal length [metres]
‘distortion’, D Distortion vector [k1 k2 k3 p1 p2]
‘distortion-bouguet’, D Distortion vector [k1 k2 p1 p2 k3]
‘canonic’ Default camera parameters: 1024 × 1024, f=1, retinal coordinates, camera at origin,
‘resolution’, N Image plane resolution: N × N or N=[W H]
‘sensor’, S Image sensor size in metres (2 × 1)
‘pixel’, S Pixel size: S × S or S=[W H]
‘color’, C Color of image plane background (default [1 1 0.8])
‘image’, IM Display an image rather than points
‘distance’, D If ‘image’ and ‘focal’ options are used, then compute pixel size based on assumed
distance D of image.
See also
Camera, FisheyeCamera, CatadioptricCamera, SphericalCamera
CentralCamera.C
Camera matrix
C = C.C() is the 3 × 4 camera matrix, also known as the camera calibration or projec-
tion matrix.
C = C.C(T) as above but for the camera at pose T (SE3).

c Corke 2017
CentralCamera.centre
Projective centre
p = C.centre() returns the 3D world coordinate of the projective centre of the camera.
Reference
Hartley & Zisserman, “Multiview Geometry”,
See also
Ray3D
CentralCamera.derivs
Compute bundle adjustment Jacobians
[p,ja,jb] = cam.derivs(T, qv, x) computes the image plane projection p (2 × 1), Jaco-
bian dP/dV (2 × 6) and Jacobian dP/dX (2 × 3) given the world point x (3 × 1) and the
camera position T (3 × 1) and orientation qv (3 × 1).
Orientation is expressed as the vector part of a unit-quaterion.
Notes
• The Jacobians are used to compute the approximate Hessian for bundle adjust-
ment problems based on camera observations of landmarks.
• This is optimized automatically generated code.
See also
UnitQuaternion, UnitQuaternion.tovec
CentralCamera.distort
Compute distorted coordinate
Xd = cam.distort(x) is the projected image plane point x (2 × 1) after lens distortion

has been applied.

c Corke 2017
CentralCamera.E
Essential matrix
E = C.E(T) is the essential matrix relating two camera views. The first view is from
the current camera pose C.T and the second is a relative motion represented by the
homogeneous transformation T.
E = C.E(C2) is the essential matrix relating two camera views described by camera
objects C (first view) and C2 (second view).
E = C.E(f) is the essential matrix based on the fundamental matrix f (3 × 3) and the
intrinsic parameters of camera C.
Reference
Y.Ma, J.Kosecka, S.Soatto, S.Sastry, “An invitation to 3D”, Springer, 2003. p.177
See also
CentralCamera.F, CentralCamera.invE
CentralCamera.estpose
Estimate pose from object model and camera view
T = C.estpose(xyz, uv) is an estimate of the pose of the object defined by coordinates

xyz (3 × N) in its own coordinate frame. uv (2 × N) are the corresponding image plane
coordinates.
Reference
“EPnP: An accurate O(n) solution to the PnP problem”, V. Lepetit, F. Moreno-Noguer,

and P. Fua, Int. Journal on Computer Vision, vol. 81, pp. 155-166, Feb. 2009.
CentralCamera.F
Fundamental matrix
F = C.F(T) is the fundamental matrix relating two camera views. The first view is
from the current camera pose C.T and the second is a relative motion represented by
the homogeneous transformation T.

c Corke 2017
F = C.F(C2) is the fundamental matrix relating two camera views described by camera
objects C (first view) and C2 (second view).
Reference
Y.Ma, J.Kosecka, S.Soatto, S.Sastry, “An invitation to 3D”, Springer, 2003. p.177
See also
CentralCamera.E
CentralCamera.flowfield
Optical flow
C.flowfield(v) displays the optical flow pattern for a sparse grid of points when the
camera has a spatial velocity v (6 × 1).
See also
quiver
CentralCamera.fov
Camera field-of-view angles.
a = C.fov() are the field of view angles (2 × 1) in radians for the camera x and y (hori-
zontal and vertical) directions.
CentralCamera.H
Homography matrix
H = C.H(T, n, d) is a 3 × 3 homography matrix for the camera observing the plane

with normal n and at distance d, from two viewpoints. The first view is from the
current camera pose C.T and the second is after a relative motion represented by the
homogeneous transformation T.

c Corke 2017
See also
CentralCamera.H
CentralCamera.invE
Decompose essential matrix
s = C.invE(E) decomposes the essential matrix E (3 × 3) into the camera motion. In

practice there are multiple solutions and s (4 × 4 × N) is a set of homogeneous trans-
formations representing possible camera motion.
s = C.invE(E, p) as above but only solutions in which the world point p is visible are
returned.
Reference
Hartley & Zisserman, “Multiview Geometry”, Chap 9, p. 259

Y.Ma, J.Kosecka, s.Soatto, s.Sastry, “An invitation to 3D”, Springer, 2003. p116, p120-
122
Notes
• The transformation is from view 1 to view 2.
See also
CentralCamera.E
CentralCamera.invH
Decompose homography matrix
s = C.invH(H) decomposes the homography H (3 × 3) into the camera motion and the
normal to the plane.
In practice there are multiple solutions and s is a vector of structures with elements:
• T, camera motion as a homogeneous transform matrix (4 × 4), translation not to
scale
• n, normal vector to the plane (3 × 3)

c Corke 2017
Notes
• There are up to 4 solutions

• Only those solutions that obey the positive depth constraint are returned
• The required camera intrinsics are taken from the camera object
• The transformation is from view 1 to view 2.
Reference
Y.Ma, J.Kosecka, s.Soatto, s.Sastry, “An invitation to 3D”, Springer, 2003. section 5.3
See also
CentralCamera.H
CentralCamera.K
Intrinsic parameter matrix
K = C.K() is the 3 × 3 intrinsic parameter matrix.
CentralCamera.normalized
Convert to normalized coordinate
cam.normalized(p) converts the image plane coordinate p (2×1) to normalized/retinal

coordinates.
See also
CentralCamera.project
CentralCamera.plot_epiline
Plot epipolar line
C.plot_epiline(f, p) plots the epipolar lines due to the fundamental matrix f and the
image points p.

c Corke 2017
C.plot_epiline(f, p, ls) as above but draw lines using the line style arguments ls.
H = C.plot_epiline(f, p) as above but return a vector of graphic handles, one per line.
CentralCamera.plot_line_tr
Plot line in theta-rho format
CentralCamera.plot_line_tr(L) plots lines on the camera’s image plane that are de-
scribed by columns of L with rows theta and rho respectively.
See also
Hough
CentralCamera.project
uv = C.project(p, options) are the image plane coordinates (2 × N) corresponding to

the world points p (3 × N).
[uv,vis] = C.project(p, options) as above but vis (S × N) is a logical matrix with ele-
ments true (1) if the point is visible, that is, it lies within the bounds of the image plane
and is in front of the camera.
L = C.project(pl, options) are the image plane homogeneous lines (3 × N) correspond-
ing to the world lines represented by a vector of Plucker objects (1 × N).
Options
the camera image plane. Temporarily overrides the current camera pose C.T.
‘objpose’, T Transform all points by the homogeneous transformation T before projecting them to
the camera image plane.
Notes
• If camera pose is a vector (1 × N), a camera trajectory, then uv (2 × N × S)

represents the sequence of projected points as the camera moves in the world.
• If object pose is a vector (1 × N), an object trajectory, then uv (2 × N × S) repre-
sents the sequence of projected points as the object moves in the world.

c Corke 2017
• Moving camera and object is not supported

• A camera or object pose sequence is not supported for the case of line projection.
• (u,v) values are set to NaN if the corresponding point is behind the camera.
See also
Camera.plot, CentralCamera.normalized, Plucker
CentralCamera.ray
3D ray for image point
R = C.ray(p) returns a vector of Ray3D objects, one for each point defined by the
columns of p.
Reference
Hartley & Zisserman, “Multiview Geometry”, p 162
See also
Ray3D
CentralCamera.visjac_e
Visual motion Jacobian for point feature
J = C.visjac_e(E, pl) is the image Jacobian (5 × 6) for the ellipse E (5 × 1) described

by u2 + E1v2 - 2E2uv + 2E3u + 2E4v + E5 = 0. The ellipse lies in the world plane pl
= (a,b,c,d) such that aX + bY + cZ + d = 0.
The Jacobian gives the rates of change of the ellipse parameters in terms of camera
spatial velocity.
Reference
B. Espiau, F. Chaumette, and P. Rives, “A New Approach to Visual Servoing in Robotics”,

IEEE Transactions on Robotics and Automation, vol. 8, pp. 313-326, June 1992.

c Corke 2017
See also
CentralCamera.visjac_p, CentralCamera.visjac_p_polar, CentralCamera.visjac_l
CentralCamera.visjac_l
Visual motion Jacobian for line feature
J = C.visjac_l(L, pl) is the image Jacobian (2N × 6) for the image plane lines L (2 ×
N). Each column of L is a line in theta-rho format, and the rows are theta and rho
respectively.
The lines all lie in the plane pl = (a,b,c,d) such that aX + bY + cZ + d = 0.
The Jacobian gives the rates of change of the line parameters in terms of camera spatial
velocity.
Reference
B. Espiau, F. Chaumette, and P. Rives, “A New Approach to Visual Servoing in Robotics”,

IEEE Transactions on Robotics and Automation, vol. 8, pp. 313-326, June 1992.
See also
CentralCamera.visjac_p, CentralCamera.visjac_p_polar, CentralCamera.visjac_e
CentralCamera.visjac_p
J = C.visjac_p(uv, z) is the image Jacobian (2N × 6) for the image plane points uv
(2 × N). The depth of the points from the camera is given by z which is a scalar for all
points, or a vector (N × 1) of depth for each point.
The Jacobian gives the image-plane point velocity in terms of camera spatial velocity.
Reference
“A tutorial on Visual Servo Control”, Hutchinson, Hager & Corke, IEEE Trans. R&A,
Vol 12(5), Oct, 1996, pp 651-670.

c Corke 2017
See also
CentralCamera.visjac_p_polar, CentralCamera.visjac_l, CentralCamera.visjac_e
CentralCamera.visjac_p_polar
J = C.visjac_p_polar(rt, z) is the image Jacobian (2N × 6) for the image plane points
rt (2 × N) described in polar form, radius and theta. The depth of the points from the
camera is given by z which is a scalar for all point, or a vector (N × 1) of depths for
each point.
The Jacobian gives the image-plane polar point coordinate velocity in terms of camera
spatial velocity.
Reference
“Combining Cartesian and polar coordinates in IBVS”, P. I. Corke, F. Spindler, and F.

Chaumette, in Proc. Int. Conf on Intelligent Robots and Systems (IROS), (St. Louis),
pp. 5962-5967, Oct. 2009.
See also
CentralCamera.visjac_p, CentralCamera.visjac_l, CentralCamera.visjac_e
chi2inv_rtb
Inverse chi-squared function
x = chi2inv_rtb(p, n) is the inverse chi-squared cdf function of n-degrees of freedom.
Notes
• only works for n=2

• uses a table lookup with around 6 figure accuracy
• an approximation to chi2inv() from the Statistics & Machine Learning Toolbox

c Corke 2017
See also
chi2inv
cie_primaries
Define CIE primary colors
p = cie_primaries() is a 3-vector with the wavelengths [m] of the CIE 1976 red, green
and blue primaries respectively.
closest
Find closest points in N-dimensional space.
k = closest(a, b) is the correspondence for N-dimensional point sets a (N × NA) and b

(N × NB). k (1 x NA) is such that the element J = k(I), that is, that the Ith column of a
is closest to the Jth column of b.
[k,d1] = closest(a, b) as above and d1(I)=|a(I)-b(J)| is the distance of the closest point.
[k,d1,d2] = closest(a, b) as above but also returns the distance to the second closest
point.
Notes
• Is a MEX file.
See also
distance

c Corke 2017
cmfrgb
RGB color matching function
The color matching function is the RGB tristimulus required to match a particular
spectral excitation.
rgb = cmfrgb(lambda) is the CIE color matching function (N × 3) for illumination at

wavelength lambda (N × 1) [m]. If lambda is a vector then each row of rgb is the
color matching function of the corresponding element of lambda.
rgb = cmfrgb(lambda, E) is the CIE color matching (1 × 3) function for an illumina-

tion spectrum E (N × 1) defined at corresponding wavelengths lambda (N × 1).
Notes
• Data from http://cvrl.ioo.ucl.ac.uk
• From Table I(5.5.3) of Wyszecki & Stiles (1982). (Table 1(5.5.3) of Wyszecki &
Stiles (1982) gives the Stiles & Burch functions in 250 cm-1 steps, while Table
I(5.5.3) of Wyszecki & Stiles (1982) gives them in interpolated 1 nm steps.)
• The Stiles & Burch 2-deg CMFs are based on measurements made on 10 ob-
servers. The data are referred to as pilot data, but probably represent the best
estimate of the 2 deg CMFs, since, unlike the CIE 2 deg functions (which were
reconstructed from chromaticity data), they were measured directly.
• These CMFs differ slightly from those of Stiles & Burch (1955). As noted in
footnote a on p. 335 of Table 1(5.5.3) of Wyszecki & Stiles (1982), the CMFs
have been "corrected in accordance with instructions given by Stiles & Burch
(1959)" and renormalized to primaries at 15500 (645.16), 19000 (526.32), and
22500 (444.44) cm-1
References
See also
cmfxyz, ccxyz

c Corke 2017
cmfxyz
matching function
The color matching function is the XYZ tristimulus required to match a particular
wavelength excitation.
xyz = cmfxyz(lambda) is the CIE xyz color matching function (N × 3) for illumination
at wavelength lambda (N × 1) [m]. If lambda is a vector then each row of xyz is the
color matching function of the corresponding element of lambda.
xyz = cmfxyz(lambda, E) is the CIE xyz color matching (1 × 3) function for an illu-
mination spectrum E (N × 1) defined at corresponding wavelengths lambda (N × 1).
Note
• CIE 1931 2-deg xyz CMFs from cvrl.ioo.ucl.ac.uk
References
See also
cmfrgb, ccxyz
col2im
Convert pixel vector to image
out = col2im(pix, imsize) is an image (H ×W × P) comprising the pixel values in pix

(N × P) with one row per pixel where N=H ×W . imsize is a 2-vector (N,M).
out = col2im(pix, im) as above but the dimensions of out are the same as im.
Notes
• The number of rows in pix must match the product of the elements of imsize.

c Corke 2017
See also
im2col
colnorm
Column-wise norm of a matrix
cn = colnorm(a) is a vector (1 × M) comprising the Euclidean norm of each column of

the matrix a (N × M).
See also
norm
colordistance
Colorspace distance
d = colordistance(im, rg) is the Euclidean distance on the rg-chromaticity plane from

coordinate rg=[r,g] to every pixel in the color image im. d is an image with the same
dimensions as im and the value of each pixel is the color space distance of the corre-
sponding pixel in im.
Notes
• The output image could be thresholded to determine color similarity.

• Note that Euclidean distance in the rg-chromaticity space does not correspond
well with human perception of color differences. Perceptually uniform spaces
such as Lab remedy this problem.
See also
colorspace

c Corke 2017
colorize
Colorize a greyscale image
out = colorize(im, mask, color) is a color image where each pixel in out is set to
the corresponding element of the greyscale image im or a specified color according
to whether the corresponding value of mask is true or false respectively. The color is
specified as a 3-vector (R,G,B).
out = colorize(im, func, color) as above but a the mask is the return value of the
function handle func applied to the image im, and returns a per-pixel logical result, eg.
@isnan.
Examples
Display image with values < 100 in blue

out = colorize(im, im<100, [0 0 1])
Display image with NaN values shown in red

out = colorize(im, @isnan, [1 0 0])
Notes
• With no output arguments the image is displayed.
See also
imono, icolor, ipixswitch
colorkmeans
Color image segmentation by clustering
L = colorkmeans(im, k, options) is a segmentation of the color image im into k

classes. The label image L has the same row and column dimension as im and each
pixel has a value in the range 0 to k-1 which indicates which cluster the correspond-
ing pixel belongs to. A k-means clustering of the chromaticity of all input pixels is
performed.
[L,C] = colorkmeans(im, k) as above but also returns the cluster centres C (k × 2)
where the Ith row is the rg-chromaticity of the Ith cluster and corresponds to the label
I. A k-means clustering of the chromaticity of all input pixels is performed.

c Corke 2017
[L,C,R] = colorkmeans(im, k) as above but also returns the residual R, the root mean
square error of all pixel chromaticities with respect to their cluster centre.
L = colorkmeans(im, C) is a segmentation of the color image im into k classes which
are defined by the cluster centres C (k × 2) in chromaticity space. Pixels are assigned
to the closest (Euclidean) centre. Since cluster centres are provided the k-means seg-
mentation step is not required.
Options
Various options are possible to choose the initial cluster centres for k-means:
‘random’ randomly choose k points from

‘spread’ randomly choose k values within the rectangle spanned by the input chromaticities.
‘pick’ interactively pick cluster centres
Notes
• The k-means clustering algorithm used in the first three forms is computationally
expensive and time consuming.
• Clustering is performed in xy-chromaticity space.
• The residual is an indication of quality of fit, low is good.
See also
rgb2xyz, kmeans
colorname
Map between color names and RGB values
rgb = colorname(name) is the rgb-tristimulus value (1 × 3) corresponding to the color

specified by the string name. If rgb is a cell-array (1×N) of names then rgb is a matrix
(N × 3) with each row being the corresponding tristimulus.
XYZ = colorname(name, ‘xyz’) as above but the XYZ-tristimulus value correspond-
ing to the color specified by the string name.
XY = colorname(name, ‘xy’) as above but the xy-chromaticity coordinates corre-
sponding to the color specified by the string name.

c Corke 2017
name = colorname(rgb) is a string giving the name of the color that is closest (Eu-
clidean) to the given rgb-tristimulus value (1 × 3). If rgb is a matrix (N × 3) then
return a cell-array (1 × N) of color names.
name = colorname(XYZ, ‘xyz’) as above but the color is the closest (Euclidean) to
the given XYZ-tristimulus value.
name = colorname(XYZ, ‘xy’) as above but the color is the closest (Euclidean) to the
given xy-chromaticity value with assumed Y=1.
Notes
• Color name may contain a wildcard, eg. “?burnt”

• Based on the standard X11 color database rgb.txt.
• Tristimulus values are in the range 0 to 1
colorseg
Color image segmentation using k-means
THIS FUNCTION IS DEPRECATED, USE COLORKMEANS INSTEAD
Notes
• deprecated. Use COLORKMEANS instead.
See also
colorkmeans
distance
Euclidean distances between sets of points
d = distance(a,b) is the Euclidean distances between L-dimensional points described

by the matrices a (L × M) and b (L × N) respectively. The distance d is M × N and
element d(I,J) is the distance between points a(I) and b(J).

c Corke 2017
Example
A = rand(400,100); B = rand(400,200);
d = distance(A,B);
Notes
• This fully vectorized (VERY FAST!)

• It computes the Euclidean distance between two vectors by:
||A-B|| = sqrt ( ||A||^2 + ||B||^2 - 2*A.B )
Author
Roland Bunschoten, University of Amsterdam, Intelligent Autonomous Systems (IAS)

group, Kruislaan 403 1098 SJ Amsterdam, tel.(+31)20-5257524, [email protected]
Last Rev: Oct 29 16:35:48 MET DST 1999, Tested: PC Matlab v5.2 and Solaris Matlab
v5.3, Thanx: Nikos Vlassis.
See also
closest
dtransform
distance transform
dt = dtransform(im, options) is the distance transform of the binary image im. The
value of each output pixel is the distance (pixels) to the closest set pixel.
Options
‘Euclidean’ use Euclidean distance (default)

‘cityblock’ use cityblock (Manhattan) distance
‘show’, T display the evolving distance transform, with a delay of T seconds between frames
See also
imorph, distancexform, DXform

c Corke 2017
e2h
Euclidean to homogeneous
H = e2h(E) is the homogeneous version (K+1 × N) of the Euclidean points E (K × N)

where each column represents one point in RK .
See also
h2e
EarthView
Image from Google maps
A concrete subclass of ImageSource that acquires images from Google maps.
Methods
grab Grab a frame from Google maps

size Size of image
Examples
Create an EarthView camera

ev = EarthView();
Zoom into QUT campus in Brisbane

ev.grab(-27.475722,153.0285, 17);
Show aerial view of Brisbane in satellite and map view

ev.grab(’brisbane’, 14)
ev.grab(’brisbane’, 14, ’map’)

c Corke 2017
Notes
• Google limit the number of map queries limit to 1000 unique (different) image
requests per viewer per day. A 403 error is returned if the daily quota is exceeded.
• Maximum size is 640 × 640 for free access, business users can get more.
• There are lots of conditions on what you can do with the images, particularly
with respect to publication. See the Google web site for details.
Author
Peter Corke, with some lines of code from from get_google_map by Val Schmidt.
See also
ImageSource
EarthView.EarthView
Create EarthView object
ev = EarthView(options)
Options
‘satellite’ Retrieve satellite image

‘map’ Retrieve map image
‘hybrid’ Retrieve satellite image with map overlay
‘scale’ Google map scale (default 18)
‘width’, W Set image width to W (default 640)
‘height’, H Set image height to H (default 640)
‘key’, S The Google maps key string
see also options for ImageSource.
Notes
• A key is required before you can use the Google Static Maps API. The key is
a long string that can be passed to the constructor or saved as an environment
variable GOOGLE_KEY. You need a Google account before you can register for
a key.

c Corke 2017
Notes
• Scale is 1 for the whole world, 20 is about as high a resolution as you can get.
See also
ImageSource, EarthView.grab
EarthView.char
Convert to string
EV.char() is a string representing the state of the EarthView object in human readable
form.
See also
EarthView.display
EarthView.grab
Grab an aerial image
im = EV.grab(lat, long, options) is an image of the Earth centred at the geographic

coordinate (lat, long).
im = EarthView.grab(lat, long, zoom, options) as above with the specified zoom.
zoom is an integer between 1 (zoom right out) to a maximum of 18-20 depending on
where in the world you are looking.
[im,E,n] = EarthView.grab(lat, long, options) as above but also returns the estimated
easting E and northing n. E and n are both matrices, the same size as im, whose
corresponding elements are the easting and northing are the coordinates of the pixel.
[im,E,n] = EarthView.grab(name, options) as above but uses a geocoding web site
to resolve the name to a location.
Options
‘satellite’ Retrieve satellite image

‘map’ Retrieve map image
‘roadmap’ Retrieve map image (synonym for ‘map’)
‘terrain’ Retrieve terrain map

c Corke 2017
‘hybrid’ Retrieve satellite image with map overlay

‘roads’ Retrieve a binary image that shows only roads, no labels. Roads are white, everything
else is black.
‘noplacenames’ Don’t show placenames.
‘noroadnames’ Don’t show roadnames.
Examples
Zoom into QUT campus in Brisbane

ev.grab(-27.475722,153.0285, 17);
Show aerial view of Brisbane in satellite and map view

ev.grab(’brisbane’, 14)
ev.grab(’brisbane’, 14, ’map’)
Notes
• If northing/easting outputs are requested the function deg2utm is required (from

MATLAB Central)
• The easting/northing is somewhat approximate, see get_google_map on MAT-

LAB Central.
• If no output argument is given the image is displayed using idisp.
edgelist
Return list of edge pixels for region
eg = edgelist(im, seed) is a list of edge pixels (2 × N) of a region in the image im

starting at edge coordinate seed=[X,Y]. The edgelist has one column per edge point
coordinate (x,y).
eg = edgelist(im, seed, direction) as above, but the direction of edge following is

specified. direction == 0 (default) means clockwise, non zero is counter-clockwise.
Note that direction is with respect to y-axis upward, in matrix coordinate frame, not
image frame.
[eg,d] = edgelist(im, seed, direction) as above but also returns a vector of edge seg-
ment directions which have values 1 to 8 representing W SW S SE E NW N NW
respectively.

c Corke 2017
Notes
• Coordinates are given assuming the matrix is an image, so the indices are always
in the form (x,y) or (column,row).
• im is a binary image where 0 is assumed to be background, non-zero is an object.
• seed must be a point on the edge of the region.
• The seed point is always the first element of the returned edgelist.
• 8-direction chain coding can give incorrect results when used with blobs founds
using 4-way connectivty.
Reference
• METHODS TO ESTIMATE AREAS AND PERIMETERS OF BLOB-LIKE

OBJECTS: A COMPARISON Luren Yang, Fritz Albregtsen, Tor Lgnnestad and
Per Grgttum IAPR Workshop on Machine Vision Applications Dec. 13-15, 1994,
Kawasaki
See also
ilabel
epidist
Distance of point from epipolar line
d = epidist(f, p1, p2) is the distance of the points p2 (2 × M) from the epipolar lines
due to points p1 (2 × N) where f (3 × 3) is a fundamental matrix relating the views
containing image points p1 and p2.
d (N × M) is the distance matrix where element d(i,j) is the distance from the point
p2(j) to the epipolar line due to point p1(i).
Author
Based on fmatrix code by, Nuno Alexandre Cid Martins, Coimbra, Oct 27, 1998, I.S.R.

c Corke 2017
See also
epiline, fmatrix
epiline
Draw epipolar lines
epiline(f, p) draws epipolar lines in current figure based on points p (2 × N) and the
fundamental matrix f (3 × 3). Points are specified by the columns of p.
epiline(f, p, ls) as above but draw lines using the line style arguments ls.
H = epiline(f, p, ls) as above but return a vector of graphic handles, one per line drawn.
See also
fmatrix, epidist
FeatureMatch
Feature correspondence object
This class represents the correspondence between two PointFeature objects. A vector
of FeatureMatch objects can represent the correspondence between sets of points.
Methods
plot Plot corresponding points

show Show summary statistics of corresponding points
ransac Determine inliers and outliers

inlier Return inlier matches
outlier Return outlier matches
subset Return a subset of matches
remove

c Corke 2017
display Display value of match

char Convert value of match to string
Properties
p1 Point coordinates in view 1 (2 × 1)

p2 Point coordinates in view 2 (2 × 1)
p Point coordinates in view 1 and 2 (4 × 1)
distance Match strength between the points
Properties of a vector of FeatureMatch objects are returned as a vector. If F is a vector

(N × 1) of FeatureMatch objects then F.p1 is a 2 × N matrix with each column the
corresponding view 1 point coordinate.
Note
• FeatureMatch is a reference object.

• FeatureMatch objects can be used in vectors and arrays
• Operates with all objects derived from PointFeature, such as ScalePointFeature,
SurfPointFeature and SiftPointFeature.
See also
PointFeature, SurfPointFeature, SiftPointFeature
FeatureMatch.FeatureMatch
Create a new FeatureMatch object
m = FeatureMatch(f1, f2, s) is a new FeatureMatch object describing a correspon-

dence between point features f1 and f2 with a strength of s.
m = FeatureMatch(f1, f2) as above but the strength is set to NaN.
Notes
• Only the coordinates of the PointFeature are kept.

c Corke 2017
See also
PointFeature, SurfPointFeature, SiftPointFeature
FeatureMatch.char
Convert to string
s = M.char() is a compact string representation of the match object. If M is a vector

then the string has multiple lines, one per element.
FeatureMatch.display
Display value
M.display() displays a compact human-readable representation of the feature pair. If

M is a vector then the elements are printed one per line.
Notes
expression is a FeatureMatch object and the command has no trailing semicolon.
See also
FeatureMatch.char
FeatureMatch.inlier
Inlier features
m2 = M.inlier() is a subset of the FeatureMatch vector M that are considered to be

inliers.
Notes
• Inliers are not determined until after RANSAC is run.

c Corke 2017
See also
FeatureMatch.outlier, FeatureMatch.ransac
FeatureMatch.outlier
Outlier features
m2 = M.outlier() is a subset of the FeatureMatch vector M that are considered to be

outliers.
Notes
• Outliers are not determined until after RANSAC is run.
See also
FeatureMatch.inlier, FeatureMatch.ransac
FeatureMatch.p
Feature point coordinate pairs
p = M.p() is a 4 × N matrix containing the feature point coordinates. Each column

contains the coordinates of a pair of corresponding points [u1,v1,u2,v2].
See also
FeatureMatch.p1, FeatureMatch.p2
FeatureMatch.p1
Feature point coordinates from view 1
p = M.p1() is a 2 × N matrix containing the feature points coordinates from view 1.

These are the (u,v) properties of the feature F1 passed to the constructor.

c Corke 2017
See also
FeatureMatch.FeatureMatch, FeatureMatch.p2, FeatureMatch.p
FeatureMatch.p2
Feature point coordinates from view 2
p = M.p2() is a 2 × N matrix containing the feature points coordinates from view 1.

These are the (u,v) properties of the feature F2 passed to the constructor.
See also
FeatureMatch.FeatureMatch, FeatureMatch.p1, FeatureMatch.p
FeatureMatch.plot
Show corresponding points
M.plot() overlays the correspondences in the FeatureMatch vector M on the current

figure. The figure must comprise views 1 and 2 side by side, for example by:
idisp({im1,im2})
m.plot()
M.plot(ls) as above but the optional line style arguments ls are passed to plot.
Notes
• Using IDISP as above adds UserData to the figure, and an error is created if this
UserData is not found.
See also
idisp

c Corke 2017
FeatureMatch.ransac
Apply RANSAC
M.ransac(func, options) applies the RANSAC algorithm to fit the point correspon-
dences to the model described by the function func. The options are passed to the
RANSAC() function. Elements of the FeatureMatch vector have their status updated
in place to indicate whether they are inliers or outliers.
Example
f1 = isurf(im1);
f2 = isurf(im2);
m = f1.match(f2);
m.ransac( @fmatrix, 1e-4);
See also
fmatrix, homography, ransac
FeatureMatch.show
Display summary statistics of the FeatureMatch vector
M.show() is a compact summary of the FeatureMatch vector M that gives the number
of matches, inliers and outliers (and their percentages).
FeatureMatch.subset
Subset of matches
m2 = M.subset(n) is a FeatureMatch vector with no more than n elements sampled

uniformly from M.

c Corke 2017
filt1d
1-dimensional rank filter
y = filt1d(x, options) is the minimum, maximum or median value (1 × N) of the vector

x (1 × N) compute over an odd length sliding window.
Options
‘max’ Compute maximum value over the window (default)

‘min’ Compute minimum value over the window
‘median’ Compute minimum value over the window
‘width’, W Width of the window (default 5)
Notes
• If the window width is even, it is incremented by one.

• The first and last elements of x are replicated so the output vector is the same
length as the input vector.
FishEyeCamera
Fish eye camera class
A concrete class a fisheye lense projection camera.

The camera coordinate system is:
0------------> u, X
|
|
| + (principal point)
|
| Z-axis is into the page.
v, Y
This camera model assumes central projection, that is, the focal point is at z=0 and the
image plane is at z=f. The image is not inverted.
Methods
project project world points to image plane

c Corke 2017




f intrinsic: focal length [metres]
rho intrinsic: pixel dimensions (2 × 1) [metres]

Notes

See also
Camera

c Corke 2017
FishEyeCamera.FishEyeCamera
Create fisheyecamera object
C = FishEyeCamera() creates a fisheye camera with canonic parameters: f=1 and

name=’canonic’.
C = FishEyeCamera(options) as above but with specified parameters.
Options

optical axis is z-axis, u- and v-axes are parallel to x- and y- axes respectively.
‘projection’, M Fisheye model: ‘equiangular’ (default), ‘sine’, ‘equisolid’, ‘stereographic’
‘k’, K Parameter for the projection model
‘resolution’, N Image plane resolution: N × N or N=[W H].
‘sensor’, S Image sensor size [metres] (2 × 1)
‘pixel’, S Pixel size: S × S or S=[W H].
Notes
• If K is not specified it is computed such that the circular imaging region maxi-
mally fills the square image plane.
See also
Camera, CentralCamera, CatadioptricCamera, SphericalCamera
FishEyeCamera.project
uv = C.project(p, options) are the image plane coordinates for the world points p.
The columns of p (3 × N) are the world points and the columns of uv (2 × N) are the
corresponding image plane points.
Options

c Corke 2017
pose C.T.
See also
CatadioprtricCamera.plot, Camera.plot
fmatrix
Estimate fundamental matrix
f = fmatrix(p1, p2, options) is the fundamental matrix (3 × 3) that relates two sets of
corresponding points p1 (2 × N) and p2 (2 × N) from two different camera views.
Notes
• The points must be corresponding, no outlier rejection is performed.

• Contains a RANSAC driver, which means it can be passed to ransac().
• f is a rank 2 matrix, that is, it is singular.
Reference
Hartley and Zisserman, ‘Multiple View Geometry in Computer Vision’, page 270.
Author
Based on fundamental matrix code by Peter Kovesi, School of Computer Science &
Software Engineering, The University of Western Australia, http://www.csse.uwa.edu.au/,
See also
ransac, homography, epiline, epidist

c Corke 2017
h2e
Homogeneous to Euclidean
E = h2e(H) is the Euclidean version (K-1 × N) of the homogeneous points H (K × N)

where each column represents one point in PK .
See also
e2h
hist2d
MEX file to compute 2-D histogram.
[h,vx,vy] = hist2d(x,y)
or
[h,vx,vy] = hist2d(x,y,[x0 dx nx],[y0 dy ny])
Inputs:
x,y data points. {x(i),y(i)} is a single data point.
x0 lowest x bin’s lower edge
dx x bin width
nx number of x bins
y0 lowest y bin’s lower edge
dy y bin width
ny number of y bins
[x0,dx,nx] and [y0,dy,ny] default = [0,1,256]
Outputs:
h histogram matrix. h(i,j) = number of data points
satisfying vx(j) <= x < vx(j+1) and vy(i) <= y < vy(i+1).
vx bin lower x-ordinates (one for each column of h)

vy bin lower y-ordinates (one for each row of h)
Notes
• Data vectors x and y must be double

c Corke 2017
Author
Michael Maurer, 7 October 1994. Copyright 1994 by Michael Maurer.
hitormiss
Hit or miss transform
H = hitormiss(im, se) is the hit-or-miss transform of the binary image im with the
structuring element se. Unlike standard morphological operations S has three possible
values: 0, 1 and don’t care (represented by NaN).
References
• Robotics, Vision & Control, Section 12.5.3, P. Corke, Springer 2011.
See also
imorph, ithin, itriplepoint, iendpoint
homline
Homogeneous line from two points
L = homline(x1, y1, x2, y2) is a vector (3 × 1) which describes a line in homogeneous

form that contains the two Euclidean points (x1,y1) and (x2,y2).
Homogeneous points X (3 × 1) on the line must satisfy L’*X = 0.
See also
plot_homline

c Corke 2017
homography
Estimate homography
H = homography(p1, p2) is the homography (3 × 3) that relates two sets of corre-

sponding points p1 (2 × N) and p2 (2 × N) from two different camera views of a planar
object.
Notes
• The points must be corresponding, no outlier rejection is performed.

• The points must be projections of points lying on a world plane
• Contains a RANSAC driver, which means it can be passed to ransac().
Author
Based on homography code by Peter Kovesi, School of Computer Science & Software
Engineering, The University of Western Australia, http://www.csse.uwa.edu.au/,
See also
ransac, invhomog, fmatrix
homtrans
Apply a homogeneous transformation
p2 = homtrans(T, p) applies the homogeneous transformation T to the points stored

columnwise in p.
• If T is in SE(2) (3 × 3) and
– p is 2 × N (2D points) they are considered Euclidean (R2 )
– p is 3 × N (2D points) they are considered projective (p2 )
• If T is in SE(3) (4 × 4) and
– p is 3 × N (3D points) they are considered Euclidean (R3 )
– p is 4 × N (3D points) they are considered projective (p3 )

c Corke 2017
tp = homtrans(T, T1) applies homogeneous transformation T to the homogeneous

transformation T1, that is tp=T*T1. If T1 is a 3-dimensional transformation then T
is applied to each plane as defined by the first two dimensions, ie. if T = N × N and
T1=N × N × M then the result is N × N × M.
Notes
• T is a homogeneous transformation defining the pose of {B} with respect to {A}.

• The points are defined with respect to frame {B} and are transformed to be with
respect to frame {A}.
See also
e2h, h2e, RTBPose.mtimes
homwarp
Warp image by an homography
out = homwarp(H, im, options) is a warp of the image im obtained by applying the
homography H to the coordinates of every input pixel.
[out,offs] = homwarp(H, im, options) as above but offs is the offset of the warped tile
out with respect to the origin of im.
Options
‘full’ output image contains all the warped pixels, but its position with respect to the input
image is given by the second return value offs.
‘extrapval’, V set unmapped pixels to this value (default NaN)
‘roi’, R output image contains the specified ROI in the input image
‘scale’, S scale the output by this factor
‘dimension’, D ensure output image is D × D
‘size’, S size of output image S=[W,H]
‘coords’, {U,V} coordinate matrices for im, each same size as im.
Notes
• The edges of the resulting output image will in general not be be vertical and
horizontal lines.

c Corke 2017
See also
homography, itrim, interp2
Hough
Hough transform class
The Hough transform is a technique for finding lines in an image using a voting scheme.
For every edge pixel in the input image a set of cells in the Hough accumulator (voting
array) are incremented.
In this version of the Hough transform lines are described by:
d = y cos(theta) + x sin(theta)
where theta is the angle the line makes to horizontal axis, and d is the perpendicular
distance between (0,0) and the line. A horizontal line has theta = 0, a vertical line has
theta = pi/2 or -pi/2.
The voting array is 2-dimensional, with columns corresponding to theta and rows cor-
responding to offset (d). Theta spans the range -pi/2 to pi/2 in Ntheta steps. Offset is
in the range -rho_max to rho_max where rho_max=max(W,H).
Methods
plot Overlay detected lines

show Display the Hough accumulator
lines Return line features
char Convert Hough parameters to string
display Display Hough parameters
Properties
Nrho Number of bins in rho direction

Ntheta Number of bins in theta direction
A The Hough accumulator (Nrho x Ntheta)
rho rho values for the centre of each bin vertically
theta Theta values for the centre of each bin horizontally
edgeThresh Threshold on relative edge pixel strength
houghThresh Threshold on relative peak strength
suppress Radius of accumulator cells cleared around peak
interpWidth Width of region used for peak interpolation

c Corke 2017
Notes
• Hough is a reference object.
See also
LineFeature
Hough.Hough
Create Hough transform object
ht = Hough(E, options) is the Hough transform of the edge image E.

For every pixel in the edge image E (H ×W ) greater than a threshold the corresponding
elements of the accumulator are incremented. By default the vote is incremented by
the edge strength but votes can be made equal with the option ‘equal’. The threshold is
determined from the maximum edge strength value x ht.edgeThresh.
Options
‘equal’ All edge pixels have equal weight, otherwise the edge pixel value is the vote strength
‘points’ Pass set of points rather than an edge image, in this case E (2 × N) is a set of N points,
or E (3 × N) is a set of N points with corresponding vote strengths as the third row
‘interpwidth’, W Interpolation width (default 3)
‘houghthresh’, T Set ht.houghThresh (default 0.5)
‘edgethresh’, T Set ht.edgeThresh (default 0.1);
‘suppress’, W Set ht.suppress (default 0)
‘nbins’, N Set number of bins, if N is scalar set Nrho=Ntheta=N, else N = [Ntheta, Nrho]. Default
400 × 401.
Hough.char
Convert to string
s = HT.char() is a compact string representation of the Hough transform parameters.

c Corke 2017
Hough.display
Display value
HT.display() displays a compact human-readable string representation of the Hough

transform parameters.
Notes
expression is a Hough object and the command has no trailing semicolon.
See also
Hough.char
Hough.lines
Find lines
L = HT.lines() is a vector of LineFeature objects that represent the dominant lines in

the Hough accumulator.
L = HT.lines(n) as above but returns no more than n LineFeature objects.
Lines are the coordinates of peaks in the Hough accumulator. The highest peak is
found, refined to subpixel precision, then all elements in an HT.suppress radius around
are zeroed so as to eliminate multiple close minima. The process is repeated for all
peaks.
The peak detection loop breaks early if the remaining peak has a strength less than
HT.houghThresh times the maximum vote value.
See also
Hough.plot, LineFeature
Hough.plot
Plot line features
HT.plot() overlays all detected lines on the current figure.

c Corke 2017
HT.plot(n) overlays a maximum of n strongest lines on the current figure.

HT.plot(n, ls) as above but the optional line style arguments ls are passed to plot.
H = HT.plot() as above but returns a vector of graphics handles for each line.
See also
Hough.lines
Hough.show
Display the Hough accumulator as image
s = HT.show() displays the Hough vote accumulator as an image using the hot col-
ormap, where ‘heat’ is proportional to the number of votes.
See also
colormap, hot
humoments
Hu moments
phi = humoments(im) is the vector (7 × 1) of Hu moment invariants for the binary

image im.
Notes
• im is assumed to be a binary image of a single connected region
Reference
M-K. Hu, Visual pattern recognition by moment invariants. IRE Trans. on Information
Theory, IT-8:pp. 179-187, 1962.

c Corke 2017
See also
npq
ianimate
Display an image sequence
ianimate(im, options) displays a greyscale image sequence im (H ×W × N) or a color

image sequence im (HxWx3xN) where N is the number of frames in the sequence,
or a cell-array of length N and the elements are either greyscale (H × W ) or color
(H ×W × 3).
ianimate(im, features, options) as above but with point features overlaid. features
(N × 1) is a cell array whose elements are vectors of feature objects for the correspond-
ing frames of im. The feature is plotted using the feature object’s plot method and
additional options are passed through to that method.
Examples
Animate image sequence:

ianimate(seq);
Animate image sequence with overlaid corner features:

c = icorner(im, ’nfeat’, 200); % computer corners
ianimate(seq, c, ’gs’); % features shown as green squares
Options
‘fps’, F set the frame rate (default 5 frames/sec)

‘loop’ endlessly loop over the sequence
‘movie’, M save the animation as a series of PNG frames in the folder M
‘npoints’, N plot no more than N features per frame (default 100)
‘only’, I display only the Ith frame from the sequence
‘title’, T displays the specified title on each frame, T is a cell array (1 × N) of strings.
Notes
• If titles are not specified the title is “frame N”

• If the ‘movie’ is used the frames can be converted to a movie using a utility like
ffmpeg, for instance:

c Corke 2017
ffmpeg -i *.png -r 5 movie.mp4
or to set the bit rate explicitly

ffmpeg -i *.png -b:v 64k movie.mp4
See also
PointFeature, iharris, isurf, idisp
ibbox
Find bounding box
box = ibbox(p) is the minimal bounding box that contains the points described by the
columns of p (2 × N).
box = ibbox(im) as above but the box minimally contains the non-zero pixels in the
image im.
Notes
• The bounding box is a 2 × 2 matrix [XMIN XMAX; YMIN YMAX].
iblobs
features
f = iblobs(im, options) is a vector of RegionFeature objects that describe each con-

nected region in the image im.
Options
‘pixelaspect’, A set pixel aspect ratio, default 1.0

‘connect’, C set connectivity, 4 (default) or 8
‘greyscale’ compute greyscale moments 0 (default) or 1
‘boundary’ compute boundary (default off)
‘area’, [A1,A2] accept only blobs with area in the interval A1 to A2
‘aspect’, [S1,S2] accept only blobs with aspect ratio in the interval S1 to S2

c Corke 2017
‘touch’, T accept only blobs that touch (1) or do not touch (0) the edge (default accept all)
‘class’, C accept only blobs of pixel value C (default all)
The RegionFeature object has many properties including:
uc centroid, horizontal coordinate

vc centroid, vertical coordinate
p centroid (uc, vc)
umin bounding box, minimum horizontal coordinate
umax bounding box, maximum horizontal coordinate
vmin bounding box, minimum vertical coordinate
vmax bounding box, maximum vertical coordinate
area the number of pixels
class the value of the pixels forming this region
label the label assigned to this region
children a list of indices of features that are children of this feature
edgepoint coordinate of a point on the perimeter
edge a list of edge points 2 × N matrix
perimeter edge length (pixels)
touch true if region touches edge of the image
a major axis length of equivalent ellipse
b minor axis length of equivalent ellipse
theta angle of major ellipse axis to horizontal axis
aspect aspect ratio b/a (always <= 1.0)
circularity 1 for a circle, less for other shapes
moments a structure containing moments of order 0 to 2
References

• METHODS TO ESTIMATE AREAS AND PERIMETERS OF BLOB-LIKE
OBJECTS: A COMPARISON Luren Yang, Fritz Albregtsen, Tor Lgnnestad and
Per Grgttum IAPR Workshop on Machine Vision Applications Dec. 13-15, 1994,
Kawasaki
• Area and perimeter measurement of blobs in discrete binary pictures. Z.Kulpa.
Comput. Graph. Image Process., 6:434-451, 1977.
Notes
• The RegionFeature objects are ordered by the raster order of the top most point
(smallest v coordinate) in each blob.
• Circularity is computed using the raw perimeter length scaled down by Kulpa’s
correction factor.

c Corke 2017
See also
RegionFeature, ilabel, idisplabel, imoments
icanny
edge detection
E = icanny(im, options) is an edge image obtained using the Canny edge detector
algorithm. Hysteresis filtering is applied to the gradient image: edge pixels > th1 are
connected to adjacent pixels > th0, those below th0 are set to zero.
Options
‘sd’, S set the standard deviation for smoothing (default 1)

‘th0’, T set the lower hysteresis threshold (default 0.1 x strongest edge)
‘th1’, T set the upper hysteresis threshold (default 0.5 x strongest edge)
Reference
• “A Computational Approach To Edge Detection”, J. Canny, IEEE Trans. Pattern

Analysis and Machine Intelligence, 8(6):679âĂŞ698, 1986.
Notes
• Produces a zero image with single pixel wide edges having non-zero values.
• Larger values correspond to stronger edges.
• If th1 is zero then no hysteresis filtering is performed.
• A color image is automatically converted to greyscale first.
See also
isobel, kdgauss

c Corke 2017
iclose
closing
out = iclose(im, se, options) is the image im after morphological closing with the
structuring element se. This is a morphological dilation followed by an erosion.
out = iclose(im, se, n, options) as above but the structuring element se is applied n
times, that is n erosions followed by n dilations.
Notes
• For binary image a closing operation can be used to eliminate small black holes
in white regions.
• Cheaper to apply a smaller structuring element multiple times than one large
one, the effective structuring element is the Minkowski sum of the structuring
element with itself n times.
• Windowing options of IMORPH can be passed. By default output image is same
size as input image.
See also
iopen, idilate, ierode, imorph
icolor
Colorize a greyscale image
C = icolor(im) is a color image C (H × W × 3)where each color plane is equal to im

(H ×W ).
C = icolor(im, color) as above but each output pixel is color (3 × 1) times the corre-
sponding element of im.
Examples
Create a color image that looks the same as the greyscale image
c = icolor(im);
each set pixel in im is set to [1 1 1] in the output.

Create an rose tinted version of the greyscale image

c Corke 2017
c = icolor(im, colorname(’pink’));
each set pixel in im is set to [0 1 1] in the output.
Notes
• Can convert a monochrome sequence (H × W × N) to a color image sequence

(HxWx3xN).
See also
imono, colorize, ipixswitch
iconcat
Concatenate images
C = iconcat(im,options) concatenates images from the cell array im.

iconcat(im,options) as above but displays the concatenated images using IDISP.
[C,u] = iconcat(im,options) as above but also returns the vector u whose elements are
the coordinates of the left (or top in vertical mode) edge of the corresponding image
within the concatenated image.
Options
‘dir’, D direction of concatenation: ‘horizontal’ (default) or ‘vertical’.

‘bgval’, B value of padding pixels (default NaN)
Examples
Horizontally concatenate three images
c = iconcat({im1, im2, im3}, ’h’);
Find the first column of each of the three images
[c,u] = iconcat({im1, im2, im3}, ’h’);
where u is a 3-vector such that im3 starts in the u(3)’rd column of c.

c Corke 2017
Notes
• The images do not have to be of the same size, and smaller images are surrounded
by background pixels which can be specified.
• Works for color or greyscale images.
• Direction can be abbreviated to first character, ‘h’ or ‘v’.
• In vertical mode all images are right justified.
• In horizontal mode all images are top justified.
See also
idisp
iconvolve
Image convolution
C = iconvolve(im, k, options) is the convolution of image im with the kernel k.

iconvolve(im, k, options) as above but display the result.
Options
‘same’ output image is same size as input image (default)

‘full’ output image is larger than the input image
‘valid’ output image is smaller than the input image, and contains only valid pixels
Notes
• If the image is color (has multiple planes) the kernel is applied to each plane,
resulting in an output image with the same number of planes.
• If the kernel has multiple planes, the image is convolved with each plane of the
kernel, resulting in an output image with the same number of planes.
• This function is a convenience wrapper for the MATLAB function CONV2.
• Works for double, uint8 or uint16 images. Image and kernel must be of the same
type and the result is of the same type.
• This function replaces iconv().

c Corke 2017
See also
conv2
icorner
Corner detector
f = icorner(im, options) is a vector of PointFeature objects describing corner features

detected in the image im. This is a non-scale space detector and by default the Harris
method is used but Shi-Tomasi and Noble are also supported.
If im is an image sequence a cell array of PointFeature vectors for the correspnding
frames of im.
The PointFeature object has many properties including:
u horizontal coordinate
v vertical coordinate
strength corner strength
descriptor corner descriptor (vector)
See PointFeature for full details
Options
‘detector’, D choose the detector where D is one of ‘harris’ (default), ‘noble’ or ‘klt’
‘sigma’, S kernel width for smoothing (default 2)
‘deriv’, D kernel for gradient (default kdgauss(2))
‘cmin’, CM minimum corner strength
‘cminthresh’, CT minimum corner strength as a fraction of maximum corner strength
‘edgegap’, E don’t return features closer than E pixels to the edge of image (default 2)
‘suppress’, R don’t return a feature closer than R pixels to an earlier feature (default 0)
‘nfeat’, N return the N strongest corners (default Inf)
‘k’, K set the value of k for the Harris detector
‘patch’, P use a P × P patch of surrounding pixel values as the feature vector. The vector has
zero mean and unit norm.
‘color’ specify that im is a color image not a sequence
Example
Compute the 100 strongest Harris features for the image

c Corke 2017
c = icorner(im, ’nfeat’, 100);
and overlay them on the image

idisp(im);
c.plot();
Notes
• Corners are processed in order from strongest to weakest.

• The function stops when:
– the corner strength drops below cmin, or
– the corner strength drops below cMinThresh x strongest corner, or
– the list of corners is exhausted
• Features are returned in descending strength order
• If im has more than 2 dimensions it is either a color image or a sequence
• If im is N × M × P it is taken as an image sequence and f is a cell array whose
elements are feature vectors for the corresponding image in the sequence.
• If im is N × M × 3 it is taken as a sequence unless the option ‘color’ is given
• If im is NxMx3xP it is taken as a sequence of color images and f is a cell ar-
ray whose elements are feature vectors for the corresponding color image in the
sequence.
• The default descriptor is a vector [Ix* Iy* Ixy*] which are the unique elements
of the structure tensor, where * denotes squared and smoothed.
• The descriptor is a vector of float types to save space
References
• “A combined corner and edge detector”, C.G. Harris and M.J. Stephens, Proc.
Fourth Alvey Vision Conf., Manchester, pp 147-151, 1988.
• “Finding corners”, J.Noble, Image and Vision Computing, vol.6, pp.121-128,
May 1988.
• “Good features to track”, J. Shi and C. Tomasi, Proc. Computer Vision and
Pattern Recognition, pp. 593-593, IEEE Computer Society, 1994.
– Robotics, Vision & Control, Section 13.3, P. Corke, Springer 2011.
See also
PointFeature, isurf

c Corke 2017
icp
Point cloud alignment
T = icp(p1, p2, options) is the homogeneous transformation that best transforms the
set of points p1 to p2 using the iterative closest point algorithm.
[T,d] = icp(p1, p2, options) as above but also returns the norm of the error between
the transformed point set p2 and p1.
Options
‘dplot’, d show the points p1 and p2 at each iteration, with a delay of d [sec].
‘plot’ show the points p1 and p2 at each iteration, with a delay of 0.5 [sec].
‘maxtheta’, T limit the change in rotation at each step to T
‘maxiter’, N stop after N iterations (default 100)
‘mindelta’, T stop when the relative change in error norm is less than T (default 0.001)
‘distthresh’, T eliminate correspondences more than T x the median distance at each iteration.
Example
Create a 3D point cloud

p1 = randn(3,20);
Transform it by an arbitrary amount

T = transl(1,2,3)*eul2tr(0.1, 0.2, 0.3)
p2 = homtrans( T, p1);
Perform icp to determine the transformation that maps p1 to p2

icp(p1, p2)
Notes
• Does not require knowledge of correspondence between the points.
– The point sets may have different numbers of points.
– Points in either set may have no corresponding point.
• Points can be 2- or 3-dimensional.
• For noisy data setting distthresh and maxtheta can help to prevent the solution
from diverging.

c Corke 2017
Reference
• “A method for registration of 3D shapes”, P.Besl and H.McKay, IEEETrans.

Pattern Anal. Mach. Intell., vol. 14, no. 2, pp. 239-256, Feb. 1992.
idecimate
an image
s = idecimate(im, m) is a decimated version of the image im whose size is reduced by

m (an integer) in both dimensions. The image is smoothed with a Gaussian kernel with
standard deviation m/2 then subsampled.
s = idecimate(im, m, sd) as above but the standard deviation of the smoothing kernel
is set to sd.
s = idecimate(im, m, []) as above but no smoothing is applied prior to decimation.
Notes
• If the image has multiple planes, each plane is decimated.

• Smoothing is used to eliminate aliasing artifacts and the standard deviation should
be chosen as a function of the maximum spatial frequency in the image.
See also
iscale, ismooth, ireplicate
idilate
Morphological dilation
out = idilate(im, se, options) is the image im after morphological dilation with the
structuring element se.
out = idilate(im, se, n, options) as above but the structuring element se is applied n
times, that is n dilations.

c Corke 2017
Options
‘border’ the border value is replicated (default)

‘none’ pixels beyond the border are not included in the window
‘trim’ output is not computed for pixels where the structuring element crosses the image
border, hence output image had reduced dimensions.
‘wrap’ the image is assumed to wrap around, left to right, top to bottom.
Notes
• Cheaper to apply a smaller structuring element multiple times than one large
• Windowing options of IMORPH can be passed.
Reference
See also
ierode, iclose, iopen, imorph
idisp
image display tool
idisp(im, options) displays an image and allows interactive investigation of pixel val-
ues, linear profiles, histograms and zooming. The image is displayed in a figure with
a toolbar across the top. If im is a cell array of images, they are first concatenated
(horizontally).
User interface
• Left clicking on a pixel will display its value in a box at the top.
• The “line” button allows two points to be specified and a new figure displays
intensity along a line between those points.
• The “histo” button displays a histogram of the pixel values in a new figure. If the
image is zoomed, the histogram is computed over only those pixels in view.

c Corke 2017
• The “zoom” button requires a left-click and drag to specify a box which defines
the zoomed view.
• The “colormap” button is displayed only for greyscale images, and is a popup
button that allows different color maps to be selected.
Options
‘nogui’ don’t display the GUI

‘noaxes’ don’t display axes on the image
‘noframe’ don’t display axes or frame on the image
‘plain’ don’t display axes, frame or GUI
‘axis’, A display the image in the axes given by handle A, the ‘nogui’ option is enforced.
‘here’ display the image in the current axes
‘title’, T put the text T in the title bar of the window
‘clickfunc’, F invoke the function handle F(x,y) on a down-click in the window
‘ncolors’, N number of colors in the color map (default 256)
‘bar’ add a color bar to the image
‘print’, F write the image to file F in EPS format
‘square’ display aspect ratio so that pixels are square
‘wide’ make figure full screen width, useful for displaying stereo pair
‘flatten’ display image planes (colors or sequence) as horizontally adjacent images
‘black’, B change black to grey level B (range 0 to 1)
‘ynormal’ y-axis increases upward, image is inverted
‘histeq’ apply histogram equalization
‘cscale’, C C is a 2-vector that specifies the grey value range that spans the colormap.
‘xydata’, XY XY is a cell array whose elements are vectors that span the x- and y-axes respectively.
‘colormap’, C set the colormap to C (N × 3)
‘grey’ color map: greyscale unsigned, zero is black, maximum value is white
‘invert’ color map: greyscale unsigned, zero is white, maximum value is black
‘signed’ color map: greyscale signed, positive is blue, negative is red, zero is black
‘invsigned’ color map: greyscale signed, positive is blue, negative is red, zero is white
‘random’ color map: random values, highlights fine structure
‘dark’ color map: greyscale unsigned, darker than ‘grey’, good for superimposed graphics
‘new’ create a new figure
Notes
• Is a wrapper around the MATLAB builtin function IMAGE. See the MATLAB
help on “Display Bit-Mapped Images” for details of color mapping.
• Color images are displayed in MATLAB true color mode: pixel triples map to
display RGB values. (0,0,0) is black, (1,1,1) is white.
• Greyscale images are displayed in indexed mode: the image pixel value is mapped
through the color map to determine the display pixel value.
• For grey scale images the minimum and maximum image values are mapped to
the first and last element of the color map, which by default (’greyscale’) is the

c Corke 2017
range black to white. To set your own scaling between displayed grey level and
pixel value use the ‘cscale’ option.
• The title of the figure window by default is the name of the variable passed in as
the image, this can’t work if the first argument is an expression.
Examples
Display 2 images side by side

idisp({im1, im2})
Display image in a subplot

subplot(211)
idisp(im, ’axis’, gca);
Call a user function when you click a pixel

idisp(im, ’clickfunc’, @(x,y) fprintf(’hello %d %d\n’, x,y))
Set a colormap, in this case a MATLAB builtin one

idisp(im, ’colormap’, cool);
Display an image which contains a map of a region, perhaps an obstacle grid, that spans
real world dimensions x, y in the range -10 to 10.
idisp(map, ’xyscale’, {[-10 10], [-10 10]});
See also
image, caxis, colormap, iconcat
idisplabel
Display an image with mask
idisplabel(im, labelimage, labels) displays only those image pixels which belong to a
specific class. im is a greyscale (H ×W ) or color (H ×W × 3) image, and labelimage
(H × W ) contains integer pixel class labels for the corresponding pixels in im. The
pixel classes to be displayed are given by labels which is either a scalar or a vector of
class labels. Non-selected pixels are displayed as white by default.
idisplabel(im, labelimage, labels, bg) as above but the grey level of the non-selected
pixels is specified by bg in the range 0 to 1 for a float image or 0 to 255 for a uint8
image..

c Corke 2017
Example
We will segment the image flowers into 7 color classes

cls = colorkemans(flowers, 7);
where the matrix cls is the same size as flowers and the elements are the corresponding
pixel class, a value in the range 1 to 7. To display pixels of class 5 we use
idisplabel(flowers, cls, 5)
and to display pixels belong to class 1 or 5 we use

idisplabel(flowers, cls, [1 5])
See also
iblobs, icolorize, colorseg
idouble
Convert integer image to double
imd = idouble(im, options) is an image with double precision elements in the range 0
to 1 corresponding to the elements of im. The integer pixels im are assumed to span
the range 0 to the maximum value of their integer class.
Options
‘single’ Return an array of single precision floats instead of doubles.

‘float’ As above.
Notes
• Works for an image with arbitrary number of dimensions, eg. a color image or
image sequence.
• There is a linear mapping (scaling) of the values of imd to im.
See also
iint, cast

c Corke 2017
iendpoint
Find end points in a binary skeleton image
out = iendpoint(im) is a binary image where pixels are set if the corresponding pixel
in the binary image im is the end point of a single-pixel wide line such as found in an
image skeleton. Computed using the hit-or-miss morphological operator.
References
• Robotics, Vision & Control, Section 12.5.3 P. Corke, Springer 2011.
See also
itriplepoint, ithin, hitormiss
ierode
Morphological erosion
out = ierode(im, se, options) is the image im after morphological erosion with the
structuring element se.
out = ierode(im, se, n, options) as above but the structuring element se is applied n
times, that is n erosions.
Options

Notes
• Cheaper to apply a smaller structuring element multiple times than one large one,
the effective structuing element is the Minkowski sum of the structuring element
with itself n times.
• Windowing options of IMORPH can be passed.

c Corke 2017
Reference
See also
idilate, iclose, iopen, imorph
igamm
correction
out = igamm(im, gamma) is a gamma corrected version of the image im. All pixels
are raised to the power gamma. Gamma encoding can be performed with gamma > 1
and decoding with gamma < 1.
out = igamm(im, ‘sRGB’) is a gamma decoded version of im using the sRGB decoding
function (JPEG images sRGB encoded).
Notes
• This function was once called igamma(), but that name taken by MATLAB
method for double class objects.
• Gamma decoding should be applied to any color image prior to colometric oper-
ations.
• The exception to this is colorspace conversion using COLORSPACE which ex-

pects RGB images to be gamma encoded.
• Gamma encoding is typically performed in a camera with gamma=0.45.
• Gamma decoding is typically performed in the display with gamma=2.2.
• For images with multiple planes the gamma correction is applied to all planes.
• For images sequences the gamma correction is applied to all elements.
• For images of type double the pixels are assumed to be in the range 0 to 1.
• For images of type int the pixels are assumed in the range 0 to the maximum
value of their class. Pixels are converted first to double, processed, then con-
verted back to the integer class.

c Corke 2017
See also
iread, colorspace
igraphseg
Graph-based image segmentation
L = igraphseg(im, k, min) is a graph-based segmentation of the color image im

(H × W × 3). L (H × W ) is an image where each element is the label assigned to
the corresponding pixel in im. k is the scale parameter, and a larger value indicates a
preference for larger regions, min is the minimum region size (pixels).
L = igraphseg(im, k, min, sigma) as above and sigma is the width of a Gaussian
which is used to initially smooth the image (default 0.5).
[L,nreg] = igraphseg(im, k, min, sigma) as above but nreg is the number of regions
found.
Example
im = iread(’58060.jpg’);
[labels,maxval] = igraphseg(im, 1500, 100, 0.5);
idisp(labels)
Reference
“Efficient graph-based image segmentation”, P. Felzenszwalb and D. Huttenlocher, Int.

Journal on Computer Vision, vol. 59, pp. 167âĂŞ181, Sept. 2004.
Notes
• Requires a color uint8 image.

• The hardwork is done by a MEX file in contrib/graphseg.
• With zero smoothing the number of regions can be massive and can crash MAT-
LAB.
Author
Pedro Felzenszwalb, 2006.

c Corke 2017
See also
ithresh, imser
ihist
Image histogram
ihist(im, options) displays the image histogram. For an image with multiple planes
the histogram of each plane is given in a separate subplot.
H = ihist(im, options) is the image histogram as a column vector. For an image with
multiple planes H is a matrix with one column per image plane.
[H,x] = ihist(im, options) as above but also returns the bin coordinates as a column
vector x.
Options
‘nbins’ number of histogram bins (default 256)

‘cdf’ compute a cumulative histogram
‘normcdf’ compute a normalized cumulative histogram, whose maximum value is one
‘sorted’ histogram but with occurrence sorted in descending magnitude order. Bin coordinates
x reflect this sorting.
Example
[h,x] = ihist(im);
bar(x,h);
[h,x] = ihist(im, ’normcdf’);

plot(x,h);
Notes
• For a uint8 image the MEX function FHIST is used (if available)
– The histogram always contains 256 bins
– The bins spans the greylevel range 0-255.
• For a floating point image the histogram spans the greylevel range 0-1.
• For floating point images all NaN and Inf values are first removed.
Machine Vision Toolbox 4.1 for MATLAB

100 Copyright Peter
c Corke 2017
See also
hist
iint
Convert image to integer class
out = iint(im) is an image with unsigned 8-bit integer elements in the range 0 to 255
corresponding to the elements of the image im.
out = iint(im, class) as above but the output pixels belong to the integer class class.
Examples
Convert double precision image to 8-bit unsigned integer

im = rand(50, 50);
out = iint(im);
Convert double precision image to 16-bit unsigned integer

im = rand(50, 50);
out = iint(im, ’uint16’);
Convert 8-bit unsigned integer image to 16-bit unsigned integer

im = randi(255, 50, 50, ’uint8’);
out = iint(im, ’uint16’);
Notes
• Works for an image with arbitrary number of dimensions, eg. a color image or
image sequence.
• If the input image is floating point (single or double) the pixel values are scaled
from an input range of [0,1] to a range spanning zero to the maximum positive
value of the output integer class.
• If the input image is an integer class then the pixels are cast to change type but
not their value.
See also
idouble

101 Copyright Peter
c Corke 2017
iisum
Sum of integral image
s = iisum(ii, u1, v1, u2, v2) is the sum of pixels in the rectangular image region defined
by its top-left (u1,v1) and bottom-right (u2,v2). ii is a precomputed integral image.
See also
intgimage
ilabel
Label an image
L = ilabel(im) is a label image that indicates connected components within the image
im (H ×W ). Each pixel in L (H ×W ) is an integer label that indicates which connected
region the corresponding pixel in im belongs to. Region labels are in the range 1 to M.
[L,m] = ilabel(im) as above but returns the value of the maximum label value.
[L,m,parents] = ilabel(im) as above but also returns region hierarchy information. The
value of parents(I) is the label of the parent, or enclosing, region of region I. A value
of 0 indicates that the region has no single enclosing region, for a binary image this
means the region touches the edge of the image, for a multilevel image it means that
the region touches more than one other region.
[L,maxlabel,parents,class] = ilabel(im) as above but also returns the class of pixels
within each region. The value of class(I) is the value of the pixels that comprise region
I.
[L,maxlabel,parents,class,edge] = ilabel(im) as above but also returns the edge-touch
status of each region. If edge(I) is 1 then region I touches edge of the image, otherwise
it does not.
Notes
• This algorithm is variously known as region labelling, connectivity analysis, con-

nected component analysis, blob labelling.
• All pixels within a region have the same value (or class).
• This is a “low level” function, IBLOBS is a higher level interface.
• Is a MEX file.

102 Copyright Peter
c Corke 2017
• The image can be binary or greyscale.
• Connectivity is only performed in 2 dimensions.
• Connectivity is performed using 4 nearest neighbours by default.
– To use 8-way connectivity pass a second argument of 8, eg. ilabel(im, 8).
– 8-way connectivity introduces ambiguities, a chequerboard is two blobs.
See also
iblobs, imoments
iline
Draw a line in an image
out = iline(im, p1, p2) is a copy of the image im with a single-pixel thick line drawn
between the points p1 and p2, each a 2-vector [U,V]. The pixels on the line are set to
1.
out = iline(im, p1, p2, v) as above but the pixels on the line are set to v.
Notes
• Uses the Bresenham algorithm.
• Only works for greyscale images.
• The line looks jagged since no anti-aliasing is performed.
See also
bresenham, iprofile, ipaste

103 Copyright Peter
c Corke 2017
im2col
Convert an image to pixel per row format
out = im2col(im) is a matrix (N × P) where each row represents a single of the image
im (H × W × P). The pixels are in image column order (ie. column 1, column 2 etc)
and there are N=W × H rows.
out = im2col(im, mask) as above but only includes pixels if:
• the corresponding element of mask (H ×W ) is non-zero
• the corresponding element of mask (N) is non-zero where N=H ×W
• the pixel index is included in the vector mask
See also
col2im
ImageSource
Abstract class for image sources
An abstract superclass for implementing image sources.
Methods

iscolor True if image is color
size Size of image
char Convert image source parameters to human readable string
display Display image source parameters in human readable form
See also
AxisWebCamera, Video, Movie

104 Copyright Peter
c Corke 2017
ImageSource.ImageSource
Image source constructor
i = ImageSource(options) is an ImageSource object that holds parameters related to

acquisition from some particular image source.
Options
‘width’, W Set image width to W

‘height’, H Set image height to H
‘int16’ Return image with int16 pixels
‘int32’ Return image with int32 pixels
‘grey’ Return image is greyscale
ImageSource.display
Display value
I.display() displays the state of the image source object in human readable form.
Notes
expression is an ImageSource object and the command has no trailing semicolon.
imatch
Template matching
xm = imatch(im1, im2, u, v, H, s) is the position of the matching subimage of im1

(template) within the image im2. The template in im1 is centred at (u,v) and its half-
width is H.

105 Copyright Peter
c Corke 2017
The template is searched for within im2 inside a rectangular region, centred at (u,v)
and whose size is a function of s. If s is a scalar the search region is [-s, s, -s, s] relative
to (u,v). More generally s is a 4-vector s=[umin, umax, vmin, vmax] relative to (u,v).
The return value is xm=[DU,DV,CC] where (DU,DV) are the u- and v-offsets relative
to (u,v) and CC is the similarity score for the best match in the search region.
[xm,score] = imatch(im1, im2, u, v, H, s) as above but also returns a matrix of match-
ing score values for each template position tested. The rows correspond to horizontal
positions of the template, and columns the vertical position. The centre element corre-
sponds to (u,v).
Example
Consider a sequence of images im(:,:,N) and we find corner points in the kth image
corners = icorner(im(:,:,k), ’nfeat’, 20);
Now, for each corner we look for the 11 × 11 patch of surrounding pixels in the next
image, by searching within a 21 × 21 region
for corner=corners
xm = imatch(im(:,:,k), im(:,:,k+1), 5, 10);

if xm(3) > 0.8
fprintf(’feature (%f,%f) moved by (%f,%f) pixels)\n’, ...
corner.u, corner.v, xm(1), xm(2) );
end
end
Notes
• Useful for tracking a template in an image sequence where im1 and im2 are
consecutive images in a template and (u,v) is the coordinate of a corner point in
im1.
• Is a MEX file.
• im1 and im2 must be the same size.
• ZNCC (zero-mean normalized cross correlation) matching is used as the simi-
larity measure. A perfect match score is 1.0 but anything above 0.8 is typically
considered to be a good match.
See also
isimilarity

106 Copyright Peter
c Corke 2017
imeshgrid
Domain matrices for image
[u,v] = imeshgrid(im) are matrices that describe the domain of image im and can be
used for the evaluation of functions over the image. u and v are the same szie as im.
The element u(v,u) = u and v(v,u) = v.
[u,v] = imeshgrid(im, n) as above but...
[u,v] = imeshgrid(w, H) as above but the domain is w × H.
[u,v] = imeshgrid(size) as above but the domain is described size which is scalar size×
size or a 2-vector [w H].
See also
meshgrid
imoments
Image moments
f = imoments(im) is a RegionFeature object that describes the greyscale moments of

the image im.
f = imoments(u, v) as above but the moments are computed from the pixel coordi-
nates given as vectors u (N × 1) and v (N × 1). All pixels are equally weighted and is
effectively a binary image.
f = imoments(u, v, w) as above but the pixels have weights given by the vector w and
is effectively a greyscale image.
Properties
The RegionFeature object has many properties including:
uc centroid, horizontal coordinate

vc centroid, vertical coordinate
area the number of pixels
theta angle of major ellipse axis to horizontal axis
shape aspect ratio b/a (always <= 1.0)

107 Copyright Peter
c Corke 2017
moments a structure containing moments of order 0 to 2, the elements are m00, m10, m01, m20,
m02, m11.
See RegionFeature help for more details.
Notes
• For a binary image the zeroth moment is the number of non-zero pixels, or its
area.
• This function does not perform connectivity it considers all non-zero pixels in
the image. If connected regions are required then use IBLOBS instead.
See also
RegionFeature, iblobs
imono
Convert color image to monochrome
out = imono(im, options) is a greyscale equivalent to the color image im.
Options
‘r601’ ITU recommendation 601 (default)

‘r709’ ITU recommendation 709
‘value’ HSV value component
Notes
• This function returns a greyscale image whether passed a color or a greyscale

image. If a greyscale image is passed it is simply returned.
• Can convert a color image sequence (HxWx3xN) to a monochrome sequence
(H ×W × N).
See also
colorize, icolor, colorspace

108 Copyright Peter
c Corke 2017
imorph
Morphological neighbourhood processing
out = imorph(im, se, op) is the image im after morphological processing with the
operator op and structuring element se.
The structuring element se is a small matrix with binary values that indicate which
elements of the template window are used in the operation.
The operation op is:
‘min’ minimum value over the structuring element

‘max’ maximum value over the structuring element
‘diff’ maximum - minimum value over the structuring element
‘plusmin’ the minimum of the pixel value and the pixelwise sum of the structuring element and
source neighbourhood.
out = imorph(im, se, op, edge) as above but performance of edge pixels can be con-
trolled. The value of edge is:

Notes
• Is a MEX file.
• Performs greyscale morphology.
• The structuring element should have an odd side length.
• For binary image ‘min’ = EROSION, ‘max’ = DILATION.
• The ‘plusmin’ operation can be used to compute the distance transform.
• The input can be logical, uint8, uint16, float or double, the output is always
double
See also
irank, ivar, hitormiss, iopen, iclose, dtransform

109 Copyright Peter
c Corke 2017
imser
Maximally stable extremal regions
label = imser(im, options) is a segmentation of the greyscale image im (H ×W ) based

on maximally stable extremal regions. label (H ×W ) is an image where each element is
the integer label assigned to the corresponding pixel in im. The labels are consecutive
integers starting at zero.
[label,nreg] = imser(im, options) as above but nreg is the number of regions found,
or one plus the maximum value of label.
Options
‘dark’ looking for dark features against a light background (default)

‘light’ looking for light features against a dark background
Example
im = iread(’castle_sign2.png’, ’grey’, ’double’);
[label,n] = imser(im, ’light’);
idisp(label)
Notes
• Is a wrapper for vl_mser, part of VLFeat (vlfeat.org), by Andrea Vedaldi and

Brian Fulkerson.
• vl_mser is a MEX file.
Reference
“Robust wide-baseline stereo from maximally stable extremal regions”, J. Matas, O.

Chum, M. Urban, and T. Pajdla, Image and Vision Computing, vol. 22, pp. 761-767,
Sept. 2004.
See also
ithresh, igraphseg

110 Copyright Peter
c Corke 2017
inormhist
Histogram normalization
out = inormhist(im) is a histogram normalized version of the image im.
Notes
• Highlights image detail in dark areas of an image.

• The histogram of the normalized image is approximately uniform, that is, all
grey levels ae equally likely to occur.
See also
ihist
intgimage
Compute integral image
out = intimage(im) is an integral image corresponding to im.

Integral images can be used for rapid computation of summations over rectangular
regions.
Examples
Create integral images for sum of pixels over rectangular regions

i = intimage(im);
Create integral images for sum of pixel squared values over rectangular regions
i = intimage(im.^2);
See also
iisum

111 Copyright Peter
c Corke 2017
invcamcal
camera calibration
c = invcamcal(C)
Decompose, or invert, a 3x4camera calibration matrix C.
The result is a camera object with the following parameters set:
f
sx, sy (with sx=1)
(u0, v0) principal point
Tcam is the homog xform of the world origin wrt camera
Since only f.sx and f.sy can be estimated we set sx = 1.

REF: Multiple View Geometry, Hartley&Zisserman, p 163-164
SEE ALSO: camera
iopen
Morphological opening
out = iopen(im, se, options) is the image im after morphological opening with the
structuring element se. This is a morphological erosion followed by dilation.
out = iopen(im, se, n, options) as above but the structuring element se is applied n
times, that is n erosions followed by n dilations.
Notes
• For binary image an opening operation can be used to eliminate small white
noise regions.
• It is cheaper to apply a smaller structuring element multiple times than one large
• Windowing options of IMORPH can be passed. By default output image is same
size as input image.
See also
iclose, idilate, ierode, imorph

112 Copyright Peter
c Corke 2017
ipad
Pad an image with constants
out = ipad(im, sides, n) is a padded version of the image im with a block of NaN
values n pixels wide on the sides of im as specified by sides.
out = ipad(im, sides, n, v) as above but pads with pixels of value v.
sides is a string containing one or more of the characters:
‘t’ top
‘b’ bottom
‘l’ left
‘r’ right
Examples
Add a band of zero pixels 20 pixels high across the top of the image:
ipad(im, ’t’, 20, 0)
Add a band of white pixels 10 pixels wide on all sides of the image:
ipad(im, ’tblr’, 10, 255)
Notes
• Not a tablet computer.
ipaste
Paste an image into an image
out = ipaste(im, im2, p, options) is the image im with the subimage im2 pasted in at
the position p=[U,V].
Options

113 Copyright Peter
c Corke 2017
‘centre’ The pasted image is centred at p, otherwise p is the top-left corner of the subimage in
im (default)
‘zero’ the coordinates of p start at zero, by default 1 is assumed
‘set’ im2 overwrites the pixels in im (default)
‘add’ im2 is added to the pixels in im
‘mean’ im2 is set to the mean of pixel values in im2 and im
Notes
• Pixels outside the pasted in region are unaffected.
See also
iline
ipixswitch
Pixelwise image merge
out = ipixswitch(mask, im1, im2) is an image where each pixel is selected from the
corresponding pixel in im1 or im2 according to the corresponding pixel values in mask.
If the element of mask is zero im1 is selected, otherwise im2 is selected.
im1 or im2 can contain a color descriptor which is one of:
• A scalar value corresponding to a greyscale
• A 3-vector corresponding to a color value
• A string containing the name of a color which is found using COLORNAME.
ipixswitch(mask, im1, im2) as above but the result is displayed.
Example
Read a uint8 image

im = iread(’lena.pgm’);
and set high valued pixels to red

a = ipixswitch(im>120, im, uint8([255 0 0]));
The result is a uint8 image since both arguments are uint8 images.
a = ipixswitch(im>120, im, [1 0 0]);
The result is a double precision image since the color specification is a double.

114 Copyright Peter
c Corke 2017
a = ipixswitch(im>120, im, ’red’);
The result is a double precision image since the result of colorname is a double preci-
sion 3-vector.
Notes
• im1, im2 and mask must all have the same number of rows and columns.
• If im1 and im2 are both greyscale then out is greyscale.
• If either of im1 and im2 are color then out is color.
• If either one image is double and one is integer then the integer image is first
converted to a double image.
See also
colorize, colorname
iprofile
Extract pixels along a line
v = iprofile(im, p1, p2) is a vector of pixel values extracted from the image im (H ×
W × P) between the points p1 (2 × 1) and p2 (2 × 1). v (N × P) has one row for each
point along the line and the row is the pixel value which will be a vector for a multi-
plane image.
[p,uv] = iprofile(im, p1, p2) as above but also returns the coordinates of the pixels
for each point along the line. Each row of uv is the pixel coordinate (u,v) for the
corresponding row of p.
Notes
• The Bresenham algorithm is used to find points along the line.
See also
bresenham, iline

115 Copyright Peter
c Corke 2017
ipyramid
Pyramidal image decomposition
out = ipyramid(im) is a pyramid decomposition of input image im using Gaussian

smoothing with standard deviation of 1. out is a cell array of images each one having
dimensions half that of the previous image. The pyramid is computed down to a non-
halvable image size.
out = ipyramid(im, sigma) as above but the Gaussian standard deviation is sigma.
out = ipyramid(im, sigma, n) as above but only n levels of the pyramid are computed.
Notes
• Works for greyscale images only.
See also
iscalespace, idecimate, ismooth
irank
Rank filter
out = irank(im, order, se) is a rank filtered version of im. Only pixels corresponding
to non-zero elements of the structuring element se are ranked and the orderth value in
rank becomes the corresponding output pixel value. The highest rank, the maximum,
is order=1.
out = irank(image, se, op, nbins) as above but the number of histogram bins can be
specified.
out = irank(image, se, op, nbins, edge) as above but the processing of edge pixels can
be controlled. The value of edge is:

‘trim’ output is not computed for pixels whose window crosses the border, hence output
image had reduced dimensions.
‘wrap’ the image is assumed to wrap around left-right, top-bottom.

116 Copyright Peter
c Corke 2017
Examples
5 × 5 median filter, 25 elements in the window, the median is the 12thn in rank
irank(im, 12, ones(5,5));
3 × 3 non-local maximum, find where a pixel is greater than its eight neighbours
se = ones(3,3); se(2,2) = 0;
im > irank(im, 1, se);
Notes

• Is a MEX file.
• The median is estimated from a histogram with nbins (default 256).
double
See also
imorph, ivar, iwindow
iread
Read image from file
im = iread() presents a file selection GUI from which the user can select an image file
which is returned as a matrix. On subsequent calls the initial folder is as set on the last
call.
im = iread([], OPTIONS) as above but allows options to be specified.
im = iread(path, options) as above but the GUI is set to the folder specified by path.
If the path is not absolute it is searched for on the MATLAB search path.
im = iread(file, options) reads the specified image file and returns a matrix. If the path
is not absolute it is searched for on MATLAB search path.
The image can be greyscale or color in any of a wide range of formats supported by the
MATLAB IMREAD function.
Wildcards are allowed in file names. If multiple files match a 3D or 4D image is
returned where the last dimension is the number of images in the sequence.

117 Copyright Peter
c Corke 2017
Options
‘uint8’ return an image with 8-bit unsigned integer pixels in the range 0 to 255
‘single’ return an image with single precision floating point pixels in the range 0 to 1.
‘double’ return an image with double precision floating point pixels in the range 0 to 1.
‘grey’ convert image to greyscale, if it’s color, using ITU rec 601
‘grey_709’ convert image to greyscale, if it’s color, using ITU rec 709
‘gamma’, G apply this gamma correction, either numeric or ‘sRGB’
‘reduce’, R decimate image by R in both dimensions
‘roi’, R apply the region of interest R to each image, where R=[umin umax; vmin vmax].
Examples
Read a color image and display it

>> im = iread(’lena.png’);
>> about im
im [uint8] : 512x512x3 (786.4 kB)
>> idisp(im);
Read a greyscale image sequence

>> im = iread(’seq/*.png’);
>> about im
im [uint8] : 512x512x9 (2.4 MB)
>> ianimate(im, ’loop’);
Notes
• A greyscale image is returned as an H ×W matrix
• A color image is returned as an H ×W × 3 matrix
• A greyscale image sequence is returned as an H ×W × N matrix where N is the

sequence length
• A color image sequence is returned as an HxWx3xN matrix where N is the se-

quence length
See also
idisp, ianimate, imono, igamma, imread, imwrite, path

118 Copyright Peter
c Corke 2017
irectify
Rectify stereo image pair
[out1,out2] = irectify(f, m, im1, im2) is a rectified pair of images corresponding to

im1 and im2. f (3 × 3) is the fundamental matrix relating the two views and m is a
FeatureMatch object containing point correspondences between the images.
[out1,out2,h1,h2] = irectify(f, m, im1, im2) as above but also returns the homogra-
phies h1 and h2 that warp im1 to out1 and im2 to out2 respectively.
Notes
• The resulting image pair are epipolar aligned, equivalent to the view if the two
original camera axes were parallel.
• Rectified images are required for dense stereo matching.
• The effect of lense distortion is not removed, use the camera calibration toolbox
to unwarp each image prior to rectification.
• The resulting images may have negative disparity.
• Some output pixels may have no corresponding input pixels and will be set to
NaN.
See also
FeatureMatch, istereo, homwarp, CentralCamera
ireplicate
Expand image
out = ireplicate(im, k) is an expanded version of the image (H ×W ) where each pixel

is replicated into a k × k tile. If im is H ×W the result is (KH)x(KW).
See also
idecimate, iscale

119 Copyright Peter
c Corke 2017
iroi
Extract region of interest
out = iroi(im,rect) is a subimage of the image im described by the rectangle rect=[umin,umax;

vmin,vmax].
out = iroi(im,C,s) as above but the region is centered at C=(U,V) and has a size s. If s
is scalar then W=H=s otherwise s=(W,H).
out = iroi(im) as above but the image is displayed and the user is prompted to adjust a
rubber band box to select the region of interest.
[out,rect] = iroi(im) as above but returns the coordinates of the selected region of
interest rect=[umin umax;vmin vmax].
[out,u,v] = iroi(im) as above but returns the range of u and v coordinates in the selected
region of interest, as vectors.
Notes
• If no output argument is specified then the result is displayed in a new window.
See also
idisp
irotate
Rotate image
out = irotate(im, angle, options) is a version of the image im that has been rotated
about its centre.
Options
‘outsize’, S set size of output image to H ×W where S=[W,H]

‘crop’ return central part of image, same size as im
‘scale’, S scale the image size by S (default 1)
‘extrapval’, V set background pixels to V (default 0)
‘smooth’, S initially smooth the image with a Gaussian of standard deviation S

120 Copyright Peter
c Corke 2017
Notes
• Rotation is defined with respect to a z-axis which is into the image.

• Counter-clockwise is a positive angle.
• The pixels in the corners of the resulting image will be undefined and set to the
‘extrapval’.
See also
iscale
isamesize
Automatic image trimming
out = isamesize(im1, im2) is an image derived from im1 that has the same dimensions
as im2. This is achieved by cropping and scaling.
out = isamesize(im1, im2, bias) as above but bias controls which part of the image is
cropped. bias=0.5 is symmetric cropping, bias<0.5 moves the crop window up or to
the left, while bias>0.5 moves the crop window down or to the right.
See also
iscale, iroi, itrim
iscale
Scale an image
out = iscale(im, s) is a version of im scaled in both directions by s which is a real

scalar. s>1 makes the image larger, s<1 makes it smaller.

121 Copyright Peter
c Corke 2017
Options
‘outsize’, s set size of out to H ×W where s=[W,H]

‘smooth’, s initially smooth image with Gaussian of standard deviation s (default 1). s=[] for no
smoothing.
See also
ireplicate, idecimate, irotate
iscalemax
Scale space maxima
f = iscalemax(L, s) is a vector of ScalePointFeature objects which are the maxima,

in space and scale, of the Laplacian of Gaussian (LoG) scale-space image sequence L
(H ×W × N). s (N × 1) is a vector of scale values corresponding to each plane of L.
If the pixels are considered as cubes in a larger volume, the maxima are those cubes
greater than all their 26 neighbours.
Notes
• Features are sorted into descending feature strength.
See also
iscalespace, ScalePointFeature
iscalespace
Scale-space image sequence
[g,L,s] = iscalespace(im, n, sigma) is a scale space image sequence of length n derived

from im (H × W ). The standard deviation of the smoothing Gaussian is sigma. At
each scale step the variance of the Gaussian increases by sigma2 . The first step in the
sequence is the original image.

122 Copyright Peter
c Corke 2017
g (H ×W × n) is the scale sequence, L (H ×W × n) is the absolute value of the Lapla-

cian of Gaussian (LoG) of the scale sequence, corresponding to each step of the se-
quence, and s (n × 1) is the vector of scales.
[g,L,s] = iscalespace(im, n) as above but sigma=1.
Examples
Create a scale-space image sequence

im = iread(’lena.png’, ’double’, ’grey’);
[G,L,s] = iscalespace(im, 50, 2);
Then find scale-space maxima, an array of ScalePointFeature objects.

f = iscalemax(L, s);
Look at the scalespace volume

slice(L, [], [], 5:10:50); shading interp
Notes
• The Laplacian is approximated by the the difference of adjacent Gaussians.
See also
iscalemax, ismooth, ilaplace, klog
iscolor
Test for color image
iscolor(im) is true (1) if im is a color image, that is, it its third dimension is equal to
three.
ishomog
Test if SE(3) homogeneous transformation matrix
ishomog(T) is true (1) if the argument T is of dimension 4 × 4 or 4 × 4 × N, else false

(0).

123 Copyright Peter
c Corke 2017
ishomog(T, ‘valid’) as above, but also checks the validity of the rotation sub-matrix.
Notes
• The first form is a fast, but incomplete, test for a transform is SE(3).
See also
isrot, ishomog2, isvec
ishomog2
Test if SE(2) homogeneous transformation matrix
ishomog2(T) is true (1) if the argument T is of dimension 3 × 3 or 3 × 3 × N, else false

(0).
ishomog2(T, ‘valid’) as above, but also checks the validity of the rotation sub-matrix.
Notes
• The first form is a fast, but incomplete, test for a transform in SE(3).
See also
ishomog, isrot2, isvec
isift
SIFT feature extractor
sf = isift(im, options) is a vector of SiftPointFeature objects representing scale and

rotationally invariant interest points in the image im.
Options

124 Copyright Peter
c Corke 2017
‘nfeat’, N set the number of features to return (default Inf)

‘suppress’, R set the suppression radius (default 0)
‘id’, V set the image_id of all features
‘Octaves’, N set the number of octaves of the DoG scale space
‘Levels’, L set the number of levels per octave of the DoG scale space (3)
‘FirstOctave’, N set the index of the first octave of the DoG scale space (0)
‘PeakThresh’, N set the peak selection threshold (0)
‘EdgeThresh’, N set the non-edge selection threshold (10)
‘NormThresh’, N set the minimum l2-norm of the descriptors before normalization. Descriptors below
the threshold are set to zero (-inf)
‘Magnif’, N set the descriptor magnification factor (3)
‘WindowSize’, N set the variance of the Gaussian window (2)
See VLFeat.org for more details.
Properties and methods
The SiftPointFeature object has many properties including:
strength feature strength
descriptor feature descriptor (128 × 1)
sigma feature scale
theta feature orientation [rad]
image_id a value passed as an option to isift
The SiftPointFeature object has many methods including:
plot Plot feature position

plot_scale Plot feature scale
distance Descriptor distance
match Match features
ncc Descriptor similarity
See SiftPointFeature and PointFeature classes for more details.
Notes
• Greyscale images only.

• If im is H × W × N it is considered to be an image sequence and F is a cell
array with N elements, each of which is the feature vectors for the corresponding
image in the sequence.
• If VLFeat is installed uses vl_feat()

125 Copyright Peter
c Corke 2017
– at least 5x faster
– does not return feature strength
– does not sort features by strength
– ‘nfeat’ option cannot be used, adjust ‘PeakThresh’ to control the number
of features
• Default MEX implementation by Andrea Vedaldi (2006).
– Features are returned in descending strength order.
• The SIFT algorithm is covered by US Patent 6,711,293 (March 23, 2004) held
by the Univerity of British Columbia.
• ISURF is a functional equivalent.
Reference
“Distinctive image features from scale-invariant keypoints”, David G. Lowe, Interna-

tional Journal of Computer Vision, 60, 2 (2004), pp. 91-110.
See also
SiftPointFeature, isurf, icorner
isimilarity
Locate template in image
s = isimilarity(T, im) is an image where each pixel is the ZNCC similarity of the
template T (M × M) to the M × M neighbourhood surrounding the corresonding input
pixel in im. s is same size as im.
s = isimilarity(T, im, metric) as above but the similarity metric is specified by the
function metric which can be any of @sad, @ssd, @ncc, @zsad, @zssd.
Example
Load an image of Wally/Waldo (the template)

T = iread(’wally.png’, ’double’);
then load an image of the crowd where he is hiding

crowd = iread(’wheres-wally.png’, ’double’);
Now search for him using the ZNCC matching measure

126 Copyright Peter
c Corke 2017
S = isimilarity(T, crowd, @zncc);
and display the similarity

idisp(S, ’colormap’, ’jet’, ’bar’)
The magnitude at each pixel indicates how well the template centred on that point
matches the surrounding pixels. The locations of the maxima are
[~,p] = peak2(S, 1, ’npeaks’, 5);
Now we can display the original scene

idisp(crowd)
and highlight the most likely places that Wally/Waldo is hiding

plot_circle(p, 30, ’fillcolor’, ’b’, ’alpha’, 0.3, ...
’edgecolor’, ’none’)
plot_point(p, ’sequence’, ’bold’, ’textsize’, 24, ...
’textcolor’, ’k’, ’Marker’, ’none’)
References
Notes
• For NCC and ZNCC the maximum in s corresponds to the most likely template
location. For SAD, SSD, ZSAD and ZSSD the minimum value corresponds to
the most likely location.
• Similarity is not computed for those pixels where the template crosses the image
boundary, and these output pixels are set to NaN.
• The ZNCC function is a MEX file and therefore the fastest
• User provided similarity metrics can be used, the function accepts two regions
and returns a scalar similarity score.
See also
imatch, sad, ssd, ncc, zsad, zssd, zncc

127 Copyright Peter
c Corke 2017
isize
Size of image
n = isize(im,d) is the size of the dth dimension of im.

[w,H] = isize(im) is the image width w and height H.
wh = isize(im) is the image size wh = [w H].
[w,H,p] = isize(im) is the image width w, height H and and number of planes p. Even
if the image has only two dimensions p will be one.
Notes
• A simple convenience wrapper on the MATLAB function SIZE.
See also
size
ismooth
Gaussian smoothing
out = ismooth(im, sigma) is the image im after convolution with a Gaussian kernel of
standard deviation sigma.
out = ismooth(im, sigma, options) as above but the options are passed to CONV2.
Options
‘full’ returns the full 2-D convolution (default)

‘same’ returns out the same size as im
‘valid’ returns the valid pixels only, those where the kernel does not exceed the bounds of the
image.
Notes
• By default (option ‘full’) the returned image is larger than the passed image.
• Smooths all planes of the input image.

128 Copyright Peter
c Corke 2017
• The Gaussian kernel has a unit volume.

• If input image is integer it is converted to float, convolved, then converted back
to integer.
See also
iconv, kgauss
isobel
Sobel edge detector
out = isobel(im) is an edge image computed using the Sobel edge operator convolved
with the image im. This is the norm of the vertical and horizontal gradients at each
pixel. The Sobel horizontal gradient kernel is:
1 |1 0 -1|
• – |2 0 -2| 8 |1 0 -1|
and the vertical gradient kernel is the transpose.
[gx,gy] = isobel(im) as above but returns the gradient images rather than the gradient
magnitude.
out = isobel(im,dx) as above but applies the kernel dx and dx’ to compute the hori-
zontal and vertical gradients respectively.
[gx,gy] = isobel(im,dx) as above but returns the gradient images rather than the gradi-
ent magnitude.
Notes
• Tends to produce quite thick edges.

• The resulting image is the same size as the input image.
• If the kernel dx is provided it can be of any size, not just 3 × 3, and could be
generated using KDGAUSS.
See also
ksobel, kdgauss, icanny, iconv

129 Copyright Peter
c Corke 2017
isrot
Test if SO(3) rotation matrix
isrot(R) is true (1) if the argument is of dimension 3 × 3 or 3 × 3 × N, else false (0).

isrot(R, ‘valid’) as above, but also checks the validity of the rotation matrix.
Notes
• A valid rotation matrix has determinant of 1.
See also
ishomog, isrot2, isvec
istereo
Stereo matching
d = istereo(left, right, range, H, options) is a disparity image computed from the

epipolar aligned stereo pair: the left image left (H × W ) and the right image right
(H × W ). d (H × W ) is the disparity and the value at each pixel is the horizontal shift
of the corresponding pixel in IML as observed in IMR. That is, the disparity d=d(v,u)
means that the pixel at right(v,u-d) is the same world point as the pixel at left(v,u).
range is the disparity search range, which can be a scalar for disparities in the range 0
to range, or a 2-vector [DMIN DMAX] for searches in the range DMIN to DMAX.
H is the half size of the matching window, which can be a scalar for N × N or a 2-vector
[N,M] for an N × M window.
[d,sim] = istereo(left, right, range, H, options) as above but returns sim which is
the same size as d and the elements are the peak matching score for the corresponding
elements of d. For the default matching metric ZNCC this varies between -1 (very bad)
to +1 (perfect).
[d,sim,dsi] = istereo(left, right, range, H, options) as above but returns dsi which is
the disparity space image (H × W × N) where N=DMAX-DMIN+1. The Ith plane is
the similarity of IML to IMR shifted to the left by DMIN+I-1.
[d,sim,p] = istereo(left, right, range, H, options) if the ‘interp’ option is given then
disparity is estimated to sub-pixel precision using quadratic interpolation. In this case
d is the interpolated disparity and p is a structure with elements A, B, dx. The interpo-
lation polynomial is s = Ad2 + Bd + C where s is the similarity score and d is disparity

130 Copyright Peter
c Corke 2017
relative to the integer disparity at which s is maximum. p.A and p.B are matrices the
same size as d whose elements are the per pixel values of the interpolation polynomial
coefficients. p.dx is the peak of the polynomial with respect to the integer disparity at
which s is maximum (in the range -0.5 to +0.5).
Options
‘metric’, M string that specifies the similarity metric to use which is one of ‘zncc’ (default), ‘ncc’,
‘ssd’ or ‘sad’.
‘interp’ enable subpixel interpolation and d contains non-integer values (default false)
‘vshift’, V move the right image V pixels vertically with respect to left.
Example
Load the left and right images

L = iread(’rocks2-l.png’, ’reduce’, 2);
R = iread(’rocks2-r.png’, ’reduce’, 2);
then compute stereo disparity and display it

d = istereo(L, R, [40, 90], 3);
idisp(d);
References
• Robotics, Vision & Control, Section 14.3, p. Corke, Springer 2011.
Notes
• Images must be greyscale.

• Disparity values pixels within a half-window dimension (H) of the edges will
not be valid and are set to NaN.
• The C term of the interpolation polynomial is not computed or returned.
• The A term is high where the disparity function has a sharp peak.
• Disparity and similarity score can be obtained from the disparity space image by
[sim,d] = max(dsi, [], 3)
See also
irectify, stdisp

131 Copyright Peter
c Corke 2017
istretch
Image normalization
out = istretch(im, options) is a normalized image in which all pixel values lie in the
range 0 to 1. That is, a linear mapping where the minimum value of im is mapped to 0
and the maximum value of im is mapped to 1.
Options
‘max’, M Pixels are mapped to the range 0 to M

‘range’, R R(1) is mapped to zero, R(2) is mapped to 1 (or max value).
Notes
• For an integer image the result is a double image in the range 0 to max value.
See also
inormhist
isurf
SURF feature extractor
sf = isurf(im, options) returns a vector of SurfPointFeature objects representing scale

and rotationally invariant interest points in the image im.
The SurfPointFeature object has many properties including:
descriptor feature descriptor (64 × 1 or 128 × 1)
sigma feature scale
Options

132 Copyright Peter
c Corke 2017
‘nfeat’, N set the number of features to return (default Inf)

‘thresh’, T set Hessian threshold. Increasing the threshold reduces the number of features com-
puted and reduces computation time.
‘octaves’, N number of octaves to process (default 5)
‘extended’ return 128-element descriptor (default 64)
‘upright’ don’t compute rotation invariance
‘suppress’, R set the suppression radius (default 0). Features are not returned if they are within R
[pixels] of an earlier (stronger) feature.
Example
Load the image

im = iread(’lena.pgm’);
Find the 10 strongest SURF features

sf = isurf(im, ’nfeat’, 10);
and overlay them on the original image as blue circles

idisp(im);
sf.plot_scale()
Notes
• Color images, or sequences, are first converted to greyscale.
• Features are returned in descending strength order
• If im is H × W × N it is considered to be an image sequence and F is a cell

array with N elements, each of which is the feature vectors for the corresponding
image in the sequence.
• Wraps an M-file implementation of OpenSurf by D. Kroon (U. Twente) or a

MEX-file OpenCV wrapper by Petter Strandmark.
• The sign of the Laplacian is not retained.
• The SURF algorithm is covered by an extensive suite of international patents

including US 8,165,401, EP 1850270 held by Toyota, KU Leuven and ETHZ.
See http://www.kooaba.com/en/plans_and_pricing/ip_licensing
Reference
“SURF: Speeded Up Robust Features”, Herbert Bay, Andreas Ess, Tinne Tuytelaars,
Luc Van Gool, Computer Vision and Image Understanding (CVIU), Vol. 110, No. 3,
pp. 346–359, 2008

133 Copyright Peter
c Corke 2017
See also
SurfPointFeature, isift, icorner
isvec
Test if vector
isvec(v) is true (1) if the argument v is a 3-vector, else false (0).

isvec(v, L) is true (1) if the argument v is a vector of length L, either a row- or column-
vector. Otherwise false (0).
Notes
• Differs from MATLAB builtin function ISVECTOR, the latter returns true for
the case of a scalar, isvec does not.
• Gives same result for row- or column-vector, ie. 3 × 1 or 1 × 3 gives true.
See also
ishomog, isrot
ithin
Morphological skeletonization
out = ithin(im) is the binary skeleton of the binary image im. Any non-zero region is
replaced by a network of single-pixel wide lines.
out = ithin(im,delay) as above but graphically displays each iteration of the skele-
tonization algorithm with a pause of delay seconds between each iteration.
References

134 Copyright Peter
c Corke 2017
See also
hitormiss, itriplepoint, iendpoint
ithresh
Interactive image threshold
ithresh(im) displays the image im in a window with a slider which adjusts the binary
threshold.
ithresh(im, T) as above but the initial threshold is set to T.
im2 = ithresh(im) as above but returns the thresholded image after the “done” button
in the GUI is pressed.
[im2,T] = ithresh(im) as above but also returns the threshold value.
Notes
• Greyscale image only.

• For a uint8 class image the slider range is 0 to 255.
• For a floating point class image the slider range is 0 to 1.0
• The GUI only displays the “done” button if output arguments are requested,
otherwise the threshold window operates independently.
See also
idisp
itrim
Trim images
This function has two different modes of functionality.

out = itrim(im, sides, n) is the image im with n pixels removed from the image sides
as specified by sides which is a string containing one or more of the characters:

135 Copyright Peter
c Corke 2017
‘t’ top
‘b’ bottom
‘l’ left
‘r’ right
[out1,out2] = itrim(im1,im2) returns the central parts of images im1 and im2 as out1
and out2 respectively. When images are rectified or warped the shapes can become
quite distorted and are embedded in rectangular images surrounded by black of NaN
values. This function crops out the central rectangular region of each. It assumes that
the undefined pixels in im1 and im2 have values of NaN. The same cropping is applied
to each input image.
[out1,out2] = itrim(im1,im2,T) as above but the threshold T in the range 0 to 1 is
used to adjust the level of cropping. The default is 0.5, a higher value will include
fewer NaN value in the result (smaller region), a lower value will include more (larger
region). A value of 0 will ensure that there are no NaN values in the returned region.
See also
homwarp, irectify
itriplepoint
Find triple points
out = itriplepoint(im) is a binary image where pixels are set if the corresponding
pixel in the binary image im is a triple point, that is where three single-pixel wide
line intersect. These are the Voronoi points in an image skeleton. Computed using the
hit-or-miss morphological operator.
References
See also
iendpoint, ithin, hitormiss

136 Copyright Peter
c Corke 2017
ivar
Pixel window statistics
out = ivar(im, se, op) is an image where each output pixel is the specified statistic over
the pixel neighbourhood indicated by the structuring element se which should have odd
side lengths. The elements in the neighbourhood corresponding to non-zero elements
in se are packed into a vector on which the required statistic is computed.
The operation op is one of:
‘var’ variance
‘kurt’ Kurtosis or peakiness of the distribution
‘skew’ skew or asymmetry of the distribution
out = ivar(im, se, op, edge) as above but performance at edge pixels can be controlled.
The value of edge is:

‘wrap’ the image is assumed to wrap around
Notes
• Is a MEX file.
double
See also
irank, iwindow

137 Copyright Peter
c Corke 2017
iwindow
Generalized spatial operator
out = iwindow(im, se, func) is an image where each pixel is the result of applying the
function func to a neighbourhood centred on the corresponding pixel in im. The neigh-
bourhood is defined by the size of the structuring element se which should have odd
side lengths. The elements in the neighbourhood corresponding to non-zero elements
in se are packed into a vector (in column order from top left) and passed to the specified
function handle func. The return value becomes the corresponding pixel value in out.
out = iwindow(image, se, func, edge) as above but performance of edge pixels can be
controlled. The value of edge is:

‘wrap’ the image is assumed to wrap around
Example
Compute the maximum value over a 5 × 5 window:

iwindow(im, ones(5,5), @max);
Compute the standard deviation over a 3 × 3 window:

iwindow(im, ones(3,3), @std);
Notes
• Is a MEX file.
• Is slow since the function func must be invoked once for every output pixel.
double
See also
ivar, irank

138 Copyright Peter
c Corke 2017
kcircle
Circular structuring element
k = kcircle(R) is a square matrix (W × W ) where W=2R+1 of zeros with a maximal

centred circular region of radius R pixels set to one.
k = kcircle(R,w) as above but the dimension of the kernel is explicitly specified.
Notes
• If R is a 2-element vector the result is an annulus of ones, and the two numbers
are interpretted as inner and outer radii.
See also
ones, ktriangle, imorph
kdgauss
Derivative of Gaussian kernel
k = kdgauss(sigma) is a 2-dimensional derivative of Gaussian kernel (W ×W ) of width

(standard deviation) sigma and centred within the matrix k whose half-width H = 3 ×
sigma and W=2 × H+1.
k = kdgauss(sigma, H) as above but the half-width is explictly specified.
Notes
• This kernel is the horizontal derivative of the Gaussian, dG/dx.

• The vertical derivative, dG/dy, is k’.
• This kernel is an effective edge detector.
See also
kgauss, kdog, klog, isobel, iconv

139 Copyright Peter
c Corke 2017
kdog
Difference of Gaussian kernel
k = kdog(sigma1) is a 2-dimensional difference of Gaussian kernel equal to KGAUSS(sigma1)

- KGAUSS(SIGMA2), where sigma1 > SIGMA2. By default SIGMA2 = 1.6*sigma1.
The kernel is centred within the matrix k whose half-width H = 3 × SIGMA and
W=2 × H+1.
k = kdog(sigma1, sigma2) as above but sigma2 is specified directly.
k = kdog(sigma1, sigma2, H) as above but the kernel half-width is specified.
Notes
• This kernel is similar to the Laplacian of Gaussian and is often used as an effi-
cient approximation.
See also
kgauss, kdgauss, klog, iconv
kgauss
Gaussian kernel
k = kgauss(sigma) is a 2-dimensional Gaussian kernel of standard deviation sigma,

and centred within the matrix k whose half-width is H=2 × sigma and W=2 × H+1.
k = kgauss(sigma, H) as above but the half-width H is specified.
Notes
• The volume under the Gaussian kernel is one.
See also
kdgauss, kdog, klog, iconv

140 Copyright Peter
c Corke 2017
klaplace
Laplacian kernel
k = klaplace() is the Laplacian kernel:

|0 1 0|
|1 -4 1|
|0 1 0|
Notes
• This kernel has an isotropic response to image gradient.
See also
ilaplace, iconv
klog
Laplacian of Gaussian kernel
k = klog(sigma) is a 2-dimensional Laplacian of Gaussian kernel of width (standard

deviation) sigma and centred within the matrix k whose half-width is H=3 × sigma,
and W=2 × H+1.
k = klog(sigma, H) as above but the half-width H is specified.
See also
kgauss, kdog, kdgauss, iconv, zcross
kmeans
K-means clustering
[L,C] = kmeans(x, k, options) is a k-means clustering of multi-dimensional data

points x (D × N) where N is the number of points, and D is the dimension. The data is

141 Copyright Peter
c Corke 2017
organized into k clusters based on Euclidean distance from cluster centres C (D × k). L
is a vector (N × 1) whose elements indicates which cluster the corresponding element
of x belongs to.
[L,C] = kmeans(x, k, ) as above but the initial clusters C0 (D × k) is given and column
I is the initial estimate of the centre of cluster I.
L = kmeans(x, C) is similar to above but the clustering step is not performed, it is
assumed to have been completed previously. C (D × k) contains the cluster centroids
and L (N × 1) indicates which cluster the corresponding element of x is closest to.
Options
‘random’ initial cluster centres are chosen randomly from the set of data points x (default)
‘spread’ initial cluster centres are chosen randomly from within the hypercube spanned by x.
‘initial’, C0 Provide initial cluster centers
Reference
“Pattern Recognition Principles”, Tou and Gonzalez, Addison-Wesley 1977, pp 94
ksobel
Sobel edge detector
k = ksobel() is the Sobel x-derivative kernel:

1/8 |1 0 -1|
|2 0 -2|
|1 0 -1|
Notes
• This kernel is an effective vertical edge detector

• The Sobel vertical derivative is k’
See also
isobel

142 Copyright Peter
c Corke 2017
ktriangle
Triangular kernel
k = ktriangle(w) is a triangular kernel within a rectangular matrix k. The dimensions

k are w × w if w is scalar or w(1) wide and w(2) high. The triangle is isocles and is full
width at the bottom row of the kernel and with its apex in the top row.
Examples
>> ktriangle(3)
ans =
|0 1 0|
|0 1 0|
|1 1 1|
See also
kcircle
lambda2rg
RGB chromaticity coordinates
rgb = lambda2rg(lambda) is the rg-chromaticity coordinate (1 × 2) for illumination at

the specific wavelength lambda [m]. If lambda is a vector (N × 1), then P (N × 2) is a
vector whose elements are the chromaticity coordinates at the corresponding elements
of lambda.
rgb = lambda2rg(lambda, E) is the rg-chromaticity coordinate (1 × 2) for an illumi-
nation spectrum E (N × 1) defined at corresponding wavelengths lambda (N × 1).
References
See also
cmfrgb, lambda2xy

143 Copyright Peter
c Corke 2017
lambda2xy
= LAMBDA2XY(LAMBDA) is the xy-chromaticity coordinate
(1 × 2) for
illumination at the specific wavelength LAMBDA [metres]. If LAMBDA is a vector

(N × 1), then P (N × 2) is a vector whose elements are the luminosity at the correspond-
ing elements of LAMBDA.
xy = lambda2xy(lambda, E) is the rg-chromaticity coordinate (1 × 2) for an illumina-
tion spectrum E (N × 1) defined at corresponding wavelengths lambda (N × 1).
References
See also
cmfxyz, lambda2rg
LineFeature
Line feature class
This class represents a line feature.
Methods
plot Plot the line segment

seglength Determine length of line segment
display Display value
char Convert value to string
Properties
rho Offset of the line

theta Orientation of the line
strength Feature strength
length Length of the line

144 Copyright Peter
c Corke 2017
Properties of a vector of LineFeature objects are returned as a vector. If L is a vector

(N × 1) of LineFeature objects then L.rho is an N × 1 vector of the rho element of each
feature.
Note
• LineFeature is a reference object.

• LineFeature objects can be used in vectors and arrays
See also
Hough, RegionFeature, PointFeature
LineFeature.LineFeature
Create a line feature object
L = LineFeature() is a line feature object with null parameters.

L = LineFeature(rho, theta, strength) is a line feature object with the specified prop-
erties. LENGTH is undefined.
L = LineFeature(rho, theta, strength, length) is a line feature object with the speci-
fied properties.
L = LineFeature(l2) is a deep copy of the line feature l2.
LineFeature.char
Convert to string
s = L.char() is a compact string representation of the line feature. If L is a vector then

the string has multiple lines, one per element.
LineFeature.display
Display value
L.display() displays a compact human-readable representation of the feature. If L is a

vector then the elements are printed one per line.

145 Copyright Peter
c Corke 2017
Notes
expression is a LineFeature object and the command has no trailing semicolon.
See also
LineFeature.char
LineFeature.plot
Plot line
L.plot() overlay the line on current plot.

L.plot(ls) as above but the optional line style arguments ls are passed to plot.
Notes
• If L is a vector then each element is plotted.
LineFeature.points
Return points on line segments
p = L.points(edge) is the set of points that lie along the line in the edge image edge
are determined.
See also
icanny
LineFeature.seglength
Compute length of line segments
The Hough transform identifies lines but cannot determine their length. This method
examines the edge pixels in the original image and determines the longest stretch of
non-zero pixels along the line.

146 Copyright Peter
c Corke 2017
l2 = L.seglength(edge, gap) is a copy of the line feature object with the property length
updated to the length of the line (pixels). Small gaps, less than gap pixels are tolerated.
l2 = L.seglength(edge) as above but the maximum allowable gap is 5 pixels.
See also
icanny
loadspectrum
Load spectrum data
s = loadspectrum(lambda, filename) is spectral data (N × D) from file filename in-

terpolated to wavelengths [metres] specified in lambda (N × 1). The spectral data can
be scalar (D=1) or vector (D>1) valued.
[s,lambda] = loadspectrum(lambda, filename) as above but also returns the passed
wavelength lambda.
Notes
• The file is assumed to have its first column as wavelength in metres, the remaind-
ing columns are linearly interpolated and returned as columns of s.
• The files are kept in the private folder inside the MVTB folder.
References
luminos
Photopic luminosity function
p = luminos(lambda) is the photopic luminosity function for the wavelengths in lambda

[m]. If lambda is a vector (N × 1), then p (N × 1) is a vector whose elements are the
luminosity at the corresponding elements of lambda.

147 Copyright Peter
c Corke 2017
Luminosity has units of lumens which are the intensity with which wavelengths are
perceived by the light-adapted human eye.
References
See also
rluminos
mkcube
Create cube
p = mkcube(s, options) is a set of points (3 × 8) that define the vertices of a cube of

side length s and centred at the origin.
[x,y,z] = mkcube(s, options) as above but return the rows of p as three vectors.
[x,y,z] = mkcube(s, ‘edge’, options) is a mesh that defines the edges of a cube.
Options
‘facepoint’ Add an extra point in the middle of each face, in this case the returned value is 3 × 14
(8 vertices + 6 face centres).
‘centre’, C The cube is centred at C (3 × 1) not the origin
‘pose’, T The pose of the cube coordinate frame is defined by the homogeneous transform T,
allowing all points in the cube to be translated or rotated.
‘edge’ Return a set of cube edges in MATLAB mesh format rather than points.
See also
cylinder, sphere

148 Copyright Peter
c Corke 2017
mkgrid
Create grid of points
p = mkgrid(d, s, options) is a set of points (3 x d2 ) that define a d × d planar grid of

points with side length s. The points are the columns of p. If d is a 2-vector the grid is
d(1)xD(2) points. If s is a 2-vector the side lengths are s(1)xS(2).
By default the grid lies in the XY plane, symmetric about the origin.
Options
‘pose’, T The pose of the grid coordinate frame is defined by the homogeneous transform T,
allowing all points in the plane to be translated or rotated.
morphdemo
Demonstrate morphology using animation
morphdemo(im, se, options) displays an animation to show the principles of the math-
ematical morphology operations dilation or erosion. Two windows are displayed side
by side, input binary image on the left and output image on the right. The structuring
element moves over the input image and is colored red if the result is zero, else blue.
Pixels in the output image are initially all grey but change to black or white as the
structuring element moves.
out = morphdemo(im, se, options) as above but returns the output image.
Options
‘dilate’ Perform morphological dilation

‘erode’ Perform morphological erosion
‘delay’ Time between animation frames (default 0.5s)
‘scale’, S Scale factor for output image (default 64)
‘movie’, M Write image frames to the folder M
Notes
• This is meant for small images, say 10 × 10 pixels.

149 Copyright Peter
c Corke 2017
See also
imorph, idilate, ierode
Movie
Class to read movie file
A concrete subclass of ImageSource that acquires images from a web camera built by
Axis Communications (www.axis.com).
Methods

size Size of image
Properties
curFrame The index of the frame just read

totalDuration The running time of the movie (seconds)
See also
ImageSource, Video
SEE ALSO: Video
Movie.Movie
Image source constructor
m = Movie(file, options) is an Movie object that returns frames from the movie file
file.
Options

150 Copyright Peter
c Corke 2017

‘scale’, S Subsample the image by S in both directions
‘skip’, S Read every Sth frame from the movie
Movie.char
Convert to string
M.char() is a string representing the state of the movie object in human readable form.
Movie.close
M.close() closes the connection to the movie.
Movie.grab
Acquire next frame from movie
im = M.grab() acquires the next image from the movie

im = M.grab(options) as above but allows the next frame to be specified.
Options
‘skip’, S Skip frames, and return current+S frame

‘frame’, F Return frame F within the movie
Notes
• If no output argument given the image is displayed using IDISP.

151 Copyright Peter
c Corke 2017
mpq
Image moments
m = mpq(im, p, q) is the PQth moment of the image im. That is, the sum of I(x,y).xp .yq .
See also
mpq_poly, npq, upq
mpq_poly
Polygon moments
m = mpq_poly(v, p, q) is the PQth moment of the polygon with vertices described by

the columns of v.
Notes
• The points must be sorted such that they follow the perimeter in sequence (counter-
clockwise).
• If the points are clockwise the moments will all be negated, so centroids will be
still be correct.
• If the first and last point in the list are the same, they are considered to be a single
vertex.
See also
mpq, npq_poly, upq_poly, Polygon
ncc
Normalized cross correlation
m = ncc(i1, i2) is the normalized cross-correlation between the two equally sized image
patches i1 and i2. The result m is a scalar in the interval -1 (non match) to 1 (perfect

152 Copyright Peter
c Corke 2017
match) that indicates similarity.
Notes
• A value of 1 indicates identical pixel patterns.

• The ncc similarity measure is invariant to scale changes in image intensity.
See also
zncc, sad, ssd, isimilarity
niblack
Adaptive thresholding
T = niblack(im, k, w2) is the per-pixel (local) threshold to apply to image im. T has
the same dimensions as im. The threshold at each pixel is a function of the mean and
standard deviation computed over a W ×W window, where W=2*w2+1.
[T,m,s] = niblack(im, k, w2) as above but returns the per-pixel mean m and standard
deviation s.
Example
t = niblack(im, -0.2, 20);
idisp(im >= t);
Notes
• This is an efficient algorithm very well suited for binarizing text.

• w2 should be chosen to be half the “size” of the features to be segmented, for
example, in text segmentation, the height of a character.
• A common choice of k=-0.2
Reference
An Introduction to Digital Image Processing, W. niblack, Prentice-Hall, 1986.

153 Copyright Peter
c Corke 2017
See also
otsu, ithresh
npq
Normalized central image moments
m = npq(im, p, q) is the PQth normalized central moment of the image im. That is
UPQ(im,p,q)/MPQ(im,0,0).
Notes
• The normalized central moments are invariant to translation and scale.
See also
npq_poly, mpq, upq
npq_poly
Normalized central polygon moments
m = npq_poly(v, p, q) is the PQth normalized central moment of the polygon with

vertices described by the columns of v.
Notes
clockwise).
still be correct.
• If the first and last point in the list are the same, they are considered as a single
vertex.
• The normalized central moments are invariant to translation and scale.

154 Copyright Peter
c Corke 2017
See also
mpq_poly, mpq, npq, upq, Polygon
numcols
Number of columns in matrix
nc = numcols(m) is the number of columns in the matrix m.
Notes
• Readable shorthand for SIZE(m,2);
See also
numrows, size
numrows
Number of rows in matrix
nr = numrows(m) is the number of rows in the matrix m.
Notes
• Readable shorthand for SIZE(m,1);
See also
numcols, size

155 Copyright Peter
c Corke 2017
OrientedScalePointFeature
ScalePointCorner feature object
A subclass of PointFeature for features with scale.
Methods

uv Return feature coordinate
Properties
scale feature scale
descriptor feature descriptor (vector)
Properties of a vector of ScalePointFeature objects are returned as a vector. If F is a

vector (N × 1) of ScalePointFeature objects then F.u is a 2 ×N matrix with each column
the corresponding point coordinate.
See also
PointFeature, OrientedScalePointFeature, SurfPointFeature, SiftPointFeature
OrientedScalePointFeature.OrientedScalePoin
Create a scale point feature object
f = OrientedScalePointFeature() is a point feature object with null parameters.

f = OrientedScalePointFeature(u, v) is a point feature object with specified coordi-
nates.
f = OrientedScalePointFeature(u, v, strength) as above but with specified strength.

156 Copyright Peter
c Corke 2017
f = OrientedScalePointFeature(u, v, strength, scale) as above but with specified fea-

ture scale.
f = OrientedScalePointFeature(u, v, strength, scale, theta) as above but with speci-
fied feature orientation.
OrientedScalePointFeature.plot
Plot feature
F.plot(options) overlay a marker to indicate feature point position and scale.

F.plot(options, ls) as above but the optional line style arguments ls are passed to plot.
If F is a vector then each element is plotted.
Options
‘circle’ Indicate scale by a circle (default)

‘clock’ Indicate scale by circle with one radial line for orientation
‘arrow’ Indicate scale and orientation by an arrow
‘disk’ Indicate scale by a translucent disk
‘color’, C Color of circle or disk (default green)
‘alpha’, A Transparency of disk, 1=opaque, 0=transparent (default 0.2)
‘scale’, S Scale factor for drawing circles and arrows.
Examples
Mark the feature coordinates with a white asterisk

f.plot(’w*’)
Mark each feature with a blue translucent disk

f.plot(’disk’, ’color’, ’b’, ’alpha’, 0.3);
Mark each feature with a green circle with a radial line to indicate orientation and with
exagerated scale
f.plot(’clock’, ’color’, ’g’, ’scale’, 2)
See also
ScalePointFeature.plot, PointFeature.plot, plot

157 Copyright Peter
c Corke 2017
otsu
Threshold selection
T = otsu(im) is an optimal threshold for binarizing an image with a bimodal intensity

histogram. T is a scalar threshold that maximizes the variance between the classes of
pixels below and above the thresold T.
Example
t = otsu(im);
idisp(im >= t);
Options
‘levels’, N Number of grey levels to use if image is float (default 256)

‘valley’, S Standard deviation for the Gaussian weighted valley emphasis option
Notes
• Performance for images with non-bimodal histograms can be quite poor.
Reference
A Threshold Selection Method from Gray-Level Histograms, N. otsu IEEE Trans. Sys-
tems, Man and Cybernetics Vol SMC-9(1), Jan 1979, pp 62-66
An improved method for image thresholding on the valley-emphasis method H-F Ng,
D. Jargalsaikhan etal Signal and Info Proc. Assocn. Annual Summit and Conf (AP-
SIPA) 2013 pp 1-4
See also
niblack, ithresh
peak
Find peaks in vector
yp = peak(y, options) are the values of the maxima in the vector y.

158 Copyright Peter
c Corke 2017
[yp,i] = peak(y, options) as above but also returns the indices of the maxima in the
vector y.
[yp,xp] = peak(y, x, options) as above but also returns the corresponding x-coordinates
of the maxima in the vector y. x is the same length as y and contains the corresponding
x-coordinates.
Options
‘npeaks’, N Number of peaks to return (default all)

‘scale’, S Only consider as peaks the largest value in the horizontal range +/- S points.
‘interp’, M Order of interpolation polynomial (default no interpolation)
‘plot’ Display the interpolation polynomial overlaid on the point data
Notes
• A maxima is defined as an element that larger than its two neighbours. The first
and last element will never be returned as maxima.
• To find minima, use peak(-V).
• The interp options fits points in the neighbourhood about the peak with an Mth
order polynomial and its peak position is returned. Typically choose M to be
even. In this case xp will be non-integer.
See also
peak2
peak2
Find peaks in a matrix
zp = peak2(z, options) are the peak values in the 2-dimensional signal z.

[zp,ij] = peak2(z, options) as above but also returns the indices of the maxima in the
matrix z. Use SUB2IND to convert these to row and column coordinates
Options
‘npeaks’, N Number of peaks to return (default all)

159 Copyright Peter
c Corke 2017
‘scale’, S Only consider as peaks the largest value in the horizontal and vertical range +/- S
points.
‘interp’ Interpolate peak (default no interpolation)
‘plot’ Display the interpolation polynomial overlaid on the point data
Notes
• A maxima is defined as an element that larger than its eight neighbours. Edges
elements will never be returned as maxima.
• To find minima, use peak2(-V).
• The interp options fits points in the neighbourhood about the peak with a paraboloid
and its peak position is returned. In this case ij will be non-integer.
See also
peak, sub2ind
pickregion
Pick a rectangular region of a figure using mouse
[p1,p2] = pickregion() initiates a rubberband box at the current click point and ani-
mates it so long as the mouse button remains down. Returns the first and last coordi-
nates in axis units.
Options
‘axis’, A The axis to select from (default current axis)

‘ls’, LS Line style for foreground line (default ‘:y’);
’bg’LS, Line style for background line (default ‘-k’);
‘width’, W Line width (default 2)
‘pressed’ Don’t wait for first button press, use current position
Notes
• Effectively a replacement for the builtin rbbox function which draws the box in
the wrong location on my Mac’s external monitor.

160 Copyright Peter
c Corke 2017
Author
Based on rubberband box from MATLAB Central written/Edited by Bob Hamans

([email protected]) 02-04-2003, in turn based on an idea of Sandra Mar-
tinka’s Rubberline.
plot_arrow
Draw an arrow in 2D or 3D
plot_arrow(p1, p2, options) draws an arrow from p1 to p2 (2 × 1 or 3 × 1).

plot_arrow(p, options) as above where the columns of p (2 × 2 or 3 × 2) define where
p=[p1 p2].
Options
• All options are passed through to arrow3.

• MATLAB colorspec such as ‘r’ or ‘b–’
See also
arrow3
plot_box
Draw a box
plot_box(b, options) draws a box defined by b=[XL XR; YL YR] on the current plot
with optional MATLAB linestyle options LS.
plot_box(x1,y1, x2,y2, options) draws a box with corners at (x1,y1) and (x2,y2), and
optional MATLAB linestyle options LS.
plot_box(’centre’, P, ‘size’, W, options) draws a box with center at P=[X,Y] and with
dimensions W=[WIDTH HEIGHT].
plot_box(’topleft’, P, ‘size’, W, options) draws a box with top-left at P=[X,Y] and with
dimensions W=[WIDTH HEIGHT].

161 Copyright Peter
c Corke 2017
plot_box(’matlab’, BOX, LS) draws box(es) as defined using the MATLAB convention
of specifying a region in terms of top-left coordinate, width and height. One box is
drawn for each row of BOX which is [xleft ytop width height].
Options
‘edgecolor’ the color of the circle’s edge, Matlab color spec

‘fillcolor’ the color of the circle’s interior, Matlab color spec
‘alpha’ transparency of the filled circle: 0=transparent, 1=solid
• For an unfilled box any standard MATLAB LineStyle such as ‘r’ or ‘b—’.
• For an unfilled box any MATLAB LineProperty options can be given such as
‘LineWidth’, 2.
• For a filled box any MATLAB PatchProperty options can be given.
Notes
• The box is added to the current plot irrespective of hold status.
• Additional options LS are MATLAB LineSpec options and are passed to PLOT.
See also
plot_poly, plot_circle, plot_ellipse
plot_circle
Draw a circle
plot_circle(C, R, options) draws a circle on the current plot with centre C=[X,Y] and
radius R. If C=[X,Y,Z] the circle is drawn in the XY-plane at height Z.
If C (2 × N) then N circles are drawn and H is N × 1. If R (1 × 1) then all circles have

the same radius or else R (1 × N) to specify the radius of each circle.
H = plot_circle(C, R, options) as above but return handles. For multiple circles H is a

vector of handles, one per circle.

162 Copyright Peter
c Corke 2017
Animation
First draw the circle and keep its graphic handle, then alter it, eg.
H = PLOT_CIRCLE(C, R)
PLOT_ELLIPSE(C, R, ’alter’, H);
Options
‘edgecolor’ the color of the circle’s edge, Matlab color spec

‘fillcolor’ the color of the circle’s interior, Matlab color spec
‘alpha’ transparency of the filled circle: 0=transparent, 1=solid
‘alter’, H alter existing circles with handle H
• For an unfilled circle any standard MATLAB LineStyle such as ‘r’ or ‘b—’.
• For an unfilled circle any MATLAB LineProperty options can be given such as
‘LineWidth’, 2.
• For a filled circle any MATLAB PatchProperty options can be given.
Notes
• The circle(s) is added to the current plot irrespective of hold status.
See also
plot_ellipse, plot_box, plot_poly
plot_ellipse
Draw an ellipse or ellipsoid
plot_ellipse(E, options) draws an ellipse or ellipsoid defined by X’EX = 0 on the

current plot, centred at the origin. E (2 × 2) for an ellipse and E (2 × 3) for an ellipsoid.
plot_ellipse(E, C, options) as above but centred at C=[X,Y]. If C=[X,Y,Z] the ellipse

is parallel to the XY plane but at height Z.
H = plot_ellipse(E, C, options) as above but return graphic handle.

163 Copyright Peter
c Corke 2017
Animation
First draw the ellipse and keep its graphic handle, then alter it, eg.
H = PLOT_ELLIPSE(E, C, ’r’)
PLOT_ELLIPSE(C, R, ’alter’, H);
Options
‘confidence’, C confidence interval, range 0 to 1

‘alter’, H alter existing ellipses with handle H
‘npoints’, N use N points to define the ellipse (default 40)
‘edgecolor’ color of the ellipse boundary edge, MATLAB color spec
‘fillcolor’ the color of the circle’s interior, MATLAB color spec
‘alpha’ transparency of the fillcolored circle: 0=transparent, 1=solid
‘shadow’ show shadows on the 3 walls of the plot box
• For an unfilled ellipse any standard MATLAB LineStyle such as ‘r’ or ‘b—’.
• For an unfilled ellipse any MATLAB LineProperty options can be given such as
‘LineWidth’, 2.
• For a filled ellipse any MATLAB PatchProperty options can be given.
Notes
• If A (2 × 2) draw an ellipse, else if A(3 × 3) draw an ellipsoid.

• The ellipse is added to the current plot irrespective of hold status.
• Shadow option only valid for ellipsoids.
• If a confidence interval is given the scaling factor is com;uted using an approxi-
mate inverse chi-squared function.
See also
plot_ellipse_inv, plot_circle, plot_box, plot_poly
plot_homline
Draw a line in homogeneous form
plot_homline(L, ls) draws a line in the current plot defined by L.X = 0 where L (3×1).
The current axis limits are used to determine the endpoints of the line. MATLAB line

164 Copyright Peter
c Corke 2017
specification ls can be set. If L (3 × N) then N lines are drawn, one per column.
H = plot_homline(L, ls) as above but returns a vector of graphics handles for the lines.
Notes
• The line(s) is added to the current plot.

• The line(s) can be drawn in 3D axes but will always lie in the xy-plane.
See also
plot_box, plot_poly, homline
plot_point
Draw a point
plot_point(p, options) adds point markers to the current plot, where p (2 × N) and
each column is the point coordinate.
Options
‘textcolor’, colspec Specify color of text

‘textsize’, size Specify size of text
‘bold’ Text in bold font.
‘printf’, {fmt, data} Label points according to printf format string and corresponding element of data
‘sequence’ Label points sequentially
‘label’, L Label for point
Additional options to PLOT can be used:

• standard MATLAB LineStyle such as ‘r’ or ‘b—’
• any MATLAB LineProperty options can be given such as ‘LineWidth’, 2.
Examples
Simple point plot

P = rand(2,4);
plot_point(P);

165 Copyright Peter
c Corke 2017
Plot points with markers

plot_point(P, ’*’);
Plot points with markers

plot_point(P, ’o’, ’MarkerFaceColor’, ’b’);
Plot points with square markers and labels 1 to 4

plot_point(P, ’sequence’, ’s’);
Plot points with circles and annotations P1 to P4

data = [1 2 4 8];
plot_point(P, ’printf’, {’ P%d’, data}, ’o’);
Notes
• The point(s) and annotations are added to the current plot.

• Points can be drawn in 3D axes but will always lie in the xy-plane.
See also
plot, text
plot_poly
Draw a polygon
plot_poly(p, options) adds a polygon defined by columns of p (2 × N), in the current

plot with default line style.
H = plot_poly(p, options) as above but processes additional options and returns a
graphics handle.
Animation
plot_poly(H, T) sets the pose of the polygon with handle H to the pose given by T
(3 × 3 or 4 × 4).
Create a polygon that can be animated, then alter it, eg.
H = PLOT_POLY(P, ’animate’, ’r’)
PLOT_POLY(H, transl(2,1,0) );

166 Copyright Peter
c Corke 2017
options
‘fillcolor’,F the color of the circle’s interior, MATLAB color spec
‘alpha’, A transparency of the filled circle: 0=transparent, 1=solid.
‘edgecolor’,E edge color
‘animate’ the polygon can be animated

‘tag’, T the polygon is created with a handle graphics tag
• For an unfilled polygon any standard MATLAB LineStyle such as ‘r’ or ‘b—’.
• For an unfilled polygon any MATLAB LineProperty options can be given such
as ‘LineWidth’, 2.
• For a filled polygon any MATLAB PatchProperty options can be given.
Notes
• If p (3 × N) the polygon is drawn in 3D

• If not filled the polygon is a line segment, otherwise it is a patch object.
• The ‘animate’ option creates an hgtransform object as a parent of the polygon,
which can be animated by the last call signature above.
• The graphics are added to the current plot.
See also
plot_box, plot_circle, patch, Polygon
plot_sphere
Draw sphere
plot_sphere(C, R, ls) draws spheres in the current plot. C is the centre of the sphere
(3 × 1), R is the radius and ls is an optional MATLAB ColorSpec, either a letter or a
3-vector.
H = plot_sphere(C, R, color) as above but returns the handle(s) for the spheres.
H = plot_sphere(C, R, color, alpha) as above but alpha specifies the opacity of the
sphere where 0 is transparant and 1 is opaque. The default is 1.

167 Copyright Peter
c Corke 2017
If C (3 × N) then N sphhere are drawn and H is N × 1. If R (1 × 1) then all spheres

have the same radius or else R (1 × N) to specify the radius of each sphere.
Example
Create four spheres

plot_sphere( mkgrid(2, 1), .2, ’b’)
and now turn on a full lighting model

lighting gouraud
light
NOTES
• The sphere is always added, irrespective of figure hold state.

• The number of vertices to draw the sphere is hardwired.
Plucker
Plucker coordinate class
Concrete class to represent a line in Plucker coordinates.
Methods
line Return Plucker line coordinates (1 × 6)

side Side operator
origin_closest origin_distance distance mindist point pp L intersect
Operators
* Multiply Plucker matrix by a general matrix

| Side operator
Notes
• This is reference class object

168 Copyright Peter
c Corke 2017
• Link objects can be used in vectors and arrays
References
• Ken Shoemake, “Ray Tracing News”, Volume 11, Number 1 http://www.realtimerendering.com/resources/RTNews/htm
Plucker.Plucker
Create Plucker object
p = Plucker(p1, p2) create a Plucker object that represents the line joining the 3D
points p1 (3 × 1) and p2 (3 × 1).
p = Plucker(’points’, p1, p2) as above.
p = Plucker(’planes’, PL1, PL2) create a Plucker object that represents the line formed
by the intersection of two planes PL1, PL2 (4 × 1).
p = Plucker(’wv’, W, V) create a Plucker object from its direction W (3 × 1) and
moment vectors V (3 × 1).
p = Plucker(’Pw’, p, W) create a Plucker object from a point p (3 × 1) and direction
vector W (3 × 1).
Plucker.char
Convert to string
s = P.char() is a string showing Plucker parameters in a compact single line format.
See also
Plucker.display
Plucker.closest
Point on line closest to given point
p = PL.closest(x) is the coordinate of a point on the line that is closest to the point x
(3 × 1).
[p,d] = PL.closest(x) as above but also returns the closest distance.

169 Copyright Peter
c Corke 2017
See also
Plucker.origin_closest
Plucker.display
Display parameters
P.display() displays the Plucker parameters in compact single line format.
Notes
expression is a Plucker object and the command has no trailing semicolon.
See also
Plucker.char
Plucker.double
Convert Plucker coordinates to real vector
PL.double() is a 6 × 1 vector comprising the moment and direction vectors.
Plucker.intersect
Line intersection
PL1.intersect(pl2) is zero if the lines intersect. It is positive if pl2 passes counter-

clockwise and negative if pl2 passes clockwise. Defined as looking in direction of
PL1
• ———>
o o
• ———>
counterclockwise clockwise

170 Copyright Peter
c Corke 2017
Plucker.intersect_plane
Line intersection with plane
x = PL.intersect_plane(p) is the point where the line intersects the plane p. Planes are
structures with a normal p.n (3 × 1) and an offset p.p (1 × 1) such that p.n x + p.p = 0.
x=[] if no intersection.
[x,T] = PL.intersect_plane(p) as above but also returns the line parameters (1 × N) at

the intersection points.
See also
Plucker.point
Plucker.intersect_volume
Line intersects plot volume
p = PL.intersect_volume(bounds, line) returns a matrix (3 × N) with columns that

indicate where the line intersects the faces of the plot volume specified in terms of
[xmin xmax ymin ymax zmin zmax]. The number of columns N is either 0 (the line is
outside the plot volume) or 2. LINE is a structure with elements .p (3 × 1) a point on
the line and .v a vector parallel to the line.
[p,T] = PL.intersect_volume(bounds, line) as above but also returns the line parame-
ters (1 × N) at the intersection points.
See also
Plucker.point
Plucker.L
Skew matrix form of the line
L = PL.L() is the Plucker matrix, a 4 × 4 skew-symmetric matrix representation of the

line.

171 Copyright Peter
c Corke 2017
Notes
• For two homogeneous points P and Q on the line, PQ’-QP’ is also skew symmet-
ric.
Plucker.line
Plucker line coordinates
P.line() is a 6-vector representation of the Plucker coordinates of the line.
See also
Plucker.v, Plucker.w
Plucker.mindist
Minimum distance between two lines
d = PL1.mindist(pl2) is the minimum distance between two Plucker lines PL1 and
pl2.
Plucker.mtimes
Plucker composition
PL * M is the product of the Plucker matrix and M (4 × N).

M * PL is the product of M (N × 4) and the Plucker matrix.
Plucker.or
Operator form of side operator
P1 | P2 is the side operator which is zero whenever the lines P1 and P2 intersect or are
parallel.

172 Copyright Peter
c Corke 2017
See also
Plucker.side
Point on line closest to the origin
p = PL.origin_closest() is the coordinate of a point on the line that is closest to the

origin.
See also
Plucker.origin_distance
Plucker.origin_distance
Smallest distance from line to the origin
p = PL.origin_distance() is the smallest distance of a point on the line to the origin.
See also
Plucker.plot
Plot a line
PL.plot(options) plots the Plucker line within the current plot volume.
PL.plot(b, options) as above but plots within the plot bounds b = [XMIN XMAX
YMIN YMAX ZMIN ZMAX].
Options
• are passed to plot3.

173 Copyright Peter
c Corke 2017
See also
plot3
Plucker.point
Point on line
p = PL.point(L) is a point on the line, where L is the parametric distance along the
line from the principal point of the line.
See also
Plucker.pp
Plucker.pp
Principal point of the line
p = PL.pp() is a point on the line.
Notes
• Same as Plucker.point(0)
See also
Plucker.point
Plucker.side
Plucker side operator
x = SIDE(p1, p2) is the side operator which is zero whenever the lines p1 and p2
intersect or are parallel.

174 Copyright Peter
c Corke 2017
See also
Plucker.or
pnmfilt
Pipe image through PNM utility
out = pnmfilt(cmd) runs the external program given by the string cmd and the output
(assumed to be PNM format) is returned as out.
out = pnmfilt(cmd, im) pipes the image im through the external program given by the
string cmd and the output is returned as out. The external program must accept and
return images in PNM format.
Examples
im = pnmfilt(’ppmforge -cloud’);
im = pnmfilt(’pnmrotate 30’, lena);
Notes
• Provides access to a large number of Unix command line utilities such as Im-
ageMagick and netpbm.
• The input image is passed as stdin, the output image is assumed to come from
stdout.
• MATLAB doesn’t support i/o to pipes so the image is written to a temporary file,
the command run to another temporary file, and that is read into MATLAB.
See also
pgmfilt, iread
PointFeature
PointCorner feature object
A superclass for image corner features.

175 Copyright Peter
c Corke 2017
Methods

Properties
Properties of a vector of PointFeature objects are returned as a vector. If F is a vec-

tor (N × 1) of PointFeature objects then F.u is a 2 × N matrix with each column the
corresponding point coordinate.
See also
ScalePointFeature, SurfPointFeature, SiftPointFeature
PointFeature.PointFeature
Create a point feature object
f = PointFeature() is a point feature object with null parameters.

f = PointFeature(u, v) is a point feature object with specified coordinates.
f = PointFeature(u, v, strength) as above but with specified strength.
PointFeature.char
Convert to string
s = F.char() is a compact string representation of the point feature. If F is a vector then

the string has multiple lines, one per element.

176 Copyright Peter
c Corke 2017
PointFeature.display
Display value
F.display() displays a compact human-readable representation of the feature. If F is a

vector then the elements are printed one per line.
Notes
expression is a PointFeature object and the command has no trailing semicolon.
See also
PointFeature.char
PointFeature.distance
Distance between feature descriptors
d = F.distance(f1) is the distance between feature descriptors, the norm of the Eu-
clidean distance.
If F is a vector then d is a vector whose elements are the distance between the corre-
sponding element of F and f1.
PointFeature.match
Match point features
m = F.match(f2, options) is a vector of FeatureMatch objects that describe candidate

matches between the two vectors of point features F and f2.
[m,C] = F.match(f2, options) as above but returns a correspodence matrix where each
row contains the indices of corresponding features in F and f2 respectively.
Options
‘thresh’, T match threshold (default 0.05)

‘median’ Threshold at the median distance
‘top’, N Take top N features

177 Copyright Peter
c Corke 2017
See also
FeatureMatch
PointFeature.ncc
Feature descriptor similarity
s = F.ncc(f1) is the similarty between feature descriptors which is a scalar in the interval
-1 to 1, where 1 is perfect match.
If F is a vector then D is a vector whose elements are the distance between the corre-
sponding element of F and f1.
PointFeature.pick
Graphically select a feature
v = F.pick() is the id of the feature closest to the point clicked by the user on a plot of
the image.
PointFeature.plot
Plot feature
F.plot() overlay a white square marker at the feature position.

F.plot(ls) as above but the optional line style arguments ls are passed to plot.
polydiff
Differentiate a polynomial
pd = polydiff(p) is a vector of coefficients of a polynomial (1 × N-1) which is the

derivative of the polynomial p (1 × N).
p = [3 2 -1];
polydiff(p)
ans =

178 Copyright Peter
c Corke 2017
6 2
See also
polyval
radgrad
Radial gradient
[gr,gt] = radgrad(im) is the radial and tangential gradient of the image im. At each
pixel the image gradient vector is resolved into the radial and tangential directions.
[gr,gt] = radgrad(im, centre) as above but the centre of the image is specified as
centre=[X,Y] rather than the centre pixel of im.
radgrad(im) as above but the result is displayed graphically.
See also
isobel
ransac
Random sample and consensus
m = ransac(func, x, T, options) is the ransac algorithm that robustly fits data x to

the model represented by the function func. ransac classifies Points that support the
model as inliers and those that do not as outliers.
x typically contains corresponding point data, one column per point pair. ransac de-
termines the subset of points (inliers) that best fit the model described by the function
func and the parameter m. T is a threshold on how well a point fits the estimated, if
the fit residual is aboe the the threshold the point is considered an outlier.
[m,in] = ransac(func, x, T, options) as above but returns the vector in of column
indices of x that describe the inlier point set.
[m,in,resid] = ransac(func, x, T, options) as above but returns the final residual of
applying func to the inlier set.

179 Copyright Peter
c Corke 2017
Options
‘maxTrials’, N maximum number of iterations (default 2000)

‘maxDataTrials’, N maximum number of attempts to select a non-degenerate data set (default 100)
Model function
out = func(R) is the function passed to RANSAC and it must accept a single argument
R which is a structure:
R.cmd the operation to perform which is either (string)

R.debug display what’s going on (logical)
R.x data to work on, N point pairs (6 × N)
R.t threshold (1 × 1)
R.theta estimated quantity to test (3 × 3)
R.misc private data (cell array)
The function return value is also a structure:
out.s sample size (1 × 1)

out.x conditioned data (2D × N)
out.misc private data (cell array)
out.inliers list of inliers (1 × m)
out.valid if data is valid for estimation (logical)
out.theta estimated quantity (3 × 3)
out.resid model fit residual (1 × 1)
The values of R.cmd are:
‘size’ out.s is the minimum number of points required to compute an estimate to out.s
‘condition’ out.x = CONDITION(R.x) condition the point data
‘decondition’ out.theta = DECONDITION(R.theta) decondition the estimated model data
‘valid’ out.valid is true if a set of points is not degenerate, that is they will produce a model.
This is used to discard random samples that do not result in useful models.
‘estimate’ [out.theta,out.resid] = EST(R.x) returns the best fit model and residual for the subset
of points R.x. If this function cannot fit a model then out.theta = []. If multiple models
are found out.theta is a cell array.
‘error’ [out.inliers,out.theta] = ERR(R.theta,R.x,T) evaluates the distance from the model(s)
R.theta to the points R.x and returns the best model out.theta and the subset of R.x
that best supports (most inliers) that model.

180 Copyright Peter
c Corke 2017
Notes
• For some algorithms (eg. fundamental matrix) it is necessary to condition the

data to improve the accuracy of model estimation. For efficiency the data is
conditioned once, and the data transform parameters are kept in the .misc ele-
ment. The inverse conditioning operation is applied to the model to transform
the estimate based on conditioned data to a model applicable to the original data.
• The functions FMATRIX and HOMOG are written so as to be callable from
RANSAC, that is, they detect a structure argument.
References
• m.A. Fishler and R.C. Boles. "Random sample concensus: A paradigm for
model fitting with applications to image analysis and automated cartography".
Comm. Assoc. Comp, Mach., Vol 24, No 6, pp 381-395, 1981
• Richard Hartley and Andrew Zisserman. "Multiple View Geometry in Computer
Vision". pp 101-113. Cambridge University Press, 2001
Author
Peter Kovesi School of Computer Science & Software Engineering The University of
Western Australia pk at csse uwa edu au http://www.csse.uwa.edu.au/ pk
See also
fmatrix, homography
Ray3D
Ray in 3D space
This object represents a ray in 3D space, defined by a point on the ray and a direction
unit-vector.
Methods
intersect Intersection of ray with plane or ray

closest Closest distance between point and ray
char Ray parameters as human readable string
display Display ray parameters in human readable form

181 Copyright Peter
c Corke 2017
Properties
P0 A point on the ray (3 × 1)

d Direction of the ray, unit vector (3 × 1)
Notes
• Ray3D objects can be used in vectors and arrays
Ray3D.Ray3D
Ray constructor
R = Ray3D(p0, d) is a new Ray3D object defined by a point on the ray p0 and a

direction vector d.
Ray3D.char
Convert to string
s = R.char() is a compact string representation of the Ray3D’s value. If R is a vector

Ray3D.closest
Closest distance between point and ray
x = R.closest(p) is the point on the ray R closest to the point p.

[x,E] = R.closest(p) as above but also returns the distance E between x and p.
Ray3D.display
Display value
R.display() displays a compact human-readable representation of the Ray3D’s value.

If R is a vector then the elements are printed one per line.

182 Copyright Peter
c Corke 2017
Notes
expression is a Ray3D object and the command has no trailing semicolon.
See also
Ray3D.char
Ray3D.intersect
Intersetion of ray with line or plane
x = R.intersect(r2) is the point on R that is closest to the ray r2. If R is a vector then
then x has multiple columns, corresponding to the intersection of R(i) with r2.
[x,E] = R.intersect(r2) as above but also returns the closest distance between the rays.
x = R.intersect(p) returns the point of intersection between the ray R and the plane
p=(a,b,c,d) where aX + bY + cZ + d = 0. If R is a vector then x has multiple columns,
corresponding to the intersection of R(i) with p.
RegionFeature
Region feature class
This class represents a region feature.
Methods
boundary Return the boundary as a list

box Return the bounding box
plot Plot the centroid
plot_boundary Plot the boundary
plot_box Plot the bounding box
plot_ellipse Plot the equivalent ellipse
pick Return the index of the blob that is clicked

183 Copyright Peter
c Corke 2017
Properties
uc* centroid, horizontal coordinate

vc* centroid, vertical coordinate
p centroid (uc, vc)
umin bounding box, minimum horizontal coordinate
umax bounding box, maximum horizontal coordinate
vmin bounding box, minimum vertical coordinate
vmax bounding box, maximum vertical coordinate
area* the number of pixels
class* the value of the pixels forming this region
label* the label assigned to this region
children a list of indices of features that are children of this feature
edgepoint coordinate of a point on the perimeter
edge a list of edge points 2 × N matrix
perimeter* edge length (pixels)
touch* true if region touches edge of the image
theta* angle of major ellipse axis to horizontal axis
aspect* aspect ratio b/a (always <= 1.0)
circularity* 1 for a circle, less for other shapes
moments a structure containing moments of order 0 to 2
bbox* the bounding box, 2 × 2 matrix [umin umax; vmin vmax]
bboxarea* bounding box area
Note
• Properties indicated with a * can be determined for a vector of RegionFeatures

and the result will be a vector of those properties (not a list) with elements cor-
responding to the original vector of RegionFeatures.
• RegionFeature is a reference object.
• RegionFeature objects can be used in vectors and arrays
• This class behaves differently to LineFeature and PointFeature when getting

properties of a vector of RegionFeature objects. For example R.u_ will be a
list not a vector.
See also
iblobs, imoments

184 Copyright Peter
c Corke 2017
RegionFeature.RegionFeature
Create a region feature object
R = RegionFeature() is a region feature object with null parameters.
RegionFeature.boundary
Boundary in polar form
[d,th] = R.boundary() is a polar representation of the boundary with respect to the

centroid. d(i) and th(i) are the distance to the boundary point and the angle respec-
tively. These vectors have 400 elements irrespective of region size.
RegionFeature.box
Return bounding box
b = R.box() is the bounding box in standard Toolbox form [xmin,xmax; ymin, ymax].
RegionFeature.char
Convert to string
s = R.char() is a compact string representation of the region feature. If R is a vector

RegionFeature.contains
Test if coordinate is contained within region bounding box
R.contains(coord) true if the coordinate COORD lies within the bounding box of the
region feature R. If R is a vector, return a vector of logical values, one per input region.

185 Copyright Peter
c Corke 2017
RegionFeature.display
Display value
R.display() is a compact string representation of the region feature. If R is a vector

then the elements are printed one per line.
Notes
• this method is invoked implicitly at the command line when the result of an
expression is a RegionFeature object and the command has no trailing semicolon.
See also
RegionFeature.char
RegionFeature.pick
Select blob from mouse click
i = R.pick() is the index of the region feature within the vector of RegionFeatures R to
which the clicked point corresponds. Since regions can overlap of be contained in other
regions, the region with the smallest area of bounding box that contains the selected
point is returned.
See also
ginput, RegionFeature.inbox
RegionFeature.plot
Plot centroid
R.plot() overlay the centroid on current plot. It is indicated with overlaid o- and x-
markers.
R.plot(ls) as above but the optional line style arguments ls are passed to plot.
If R is a vector then each element is plotted.

186 Copyright Peter
c Corke 2017
RegionFeature.plot_boundary
plot boundary
R.plot_boundary() overlay perimeter points on current plot.
R.plot_boundary(ls) as above but the optional line style arguments ls are passed to
plot.
Notes
• If R is a vector then each element is plotted.
See also
boundmatch
RegionFeature.plot_box
Plot bounding box
R.plot_box() overlay the the bounding box of the region on current plot.
R.plot_box(ls) as above but the optional line style arguments ls are passed to plot.
RegionFeature.plot_ellipse
Plot equivalent ellipse
R.plot_ellipse() overlay the the equivalent ellipse of the region on current plot.
R.plot_ellipse(ls) as above but the optional line style arguments ls are passed to plot.

187 Copyright Peter
c Corke 2017
rg_addticks
Label spectral locus
rg_addticks() adds wavelength ticks to the spectral locus.
See also
xycolourspace
rgb2xyz
RGB to XYZ color space
[x, y, z] = rgb2xyz(r, g, b) xyz = rgb2xyz(rgb)

convert (R,g,b) coordinates to (X,Y,Z) color space. If RGB (or R, g, b) have more
than one row, then computation is
done row wise.
SEE ALSO: ccxyz cmfxyz
rluminos
Relative photopic luminosity function
p = rluminos(lambda) is the relative photopic luminosity function for the wavelengths

in lambda [m]. If lambda is a vector (N ×1), then p (N ×1) is a vector whose elements
are the luminosity at the corresponding elements of lambda.
Relative luminosity lies in the interval 0 to 1 which indicate the intensity with which
wavelengths are perceived by the light-adapted human eye.
References

188 Copyright Peter
c Corke 2017
See also
luminos
sad
Sum of absolute differences
m = sad(i1, i2) is the sum of absolute differences between the two equally sized image
patches i1 and i2. The result m is a scalar that indicates image similarity, a value of
0 indicates identical pixel patterns and is increasingly positive as image dissimilarity
increases.
See also
zsad, ssd, ncc, isimilarity
ScalePointFeature
ScalePointCorner feature object
A subclass of PointFeature for features with scale.
Methods

Properties

189 Copyright Peter
c Corke 2017

scale feature scale
Properties of a vector of ScalePointFeature objects are returned as a vector. If F is a

vector (N × 1) of ScalePointFeature objects then F.u is a 2 ×N matrix with each column
the corresponding point coordinate.
See also
PointFeature, OrientedScalePointFeature, SurfPointFeature, SiftPointFeature
ScalePointFeature.ScalePointFeature
Create a scale point feature object
f = ScalePointFeature() is a point feature object with null parameters.

f = ScalePointFeature(u, v) is a point feature object with specified coordinates.
f = ScalePointFeature(u, v, strength) as above but with specified strength.
f = ScalePointFeature(u, v, strength, scale) as above but with specified feature scale.
ScalePointFeature.plot
Plot feature
F.plot(options) overlay a marker at the feature position. The default is a point marker.
F.plot(options, ls) as above but the optional line style arguments ls are passed to plot.
Options
‘circle’ Indicate scale by a circle

‘disk’ Indicate scale by a translucent disk
‘color’, C Color of circle or disk (default green)
‘alpha’, A Transparency of disk, 1=opaque, 0=transparent (default 0.2)
‘scale’, S Scale factor for drawing circles and arrows.

190 Copyright Peter
c Corke 2017
Examples
Mark the feature coordinates with a white asterisk

f.plot(’w*’)
Mark each feature with a blue translucent disk

f.plot(’disk’, ’color’, ’b’, ’alpha’, 0.3);
Mark each feature with a green circle and with exagerated scale
f.plot(’circle’, ’color’, ’g’, ’scale’, 2)
See also
PointFeature.plot, plot
showcolorspace
Display spectral locus
SHOWCOLORSPACE(’xy’) display a fully colored spectral locus in terms of CIE x

and y coordinates.
SHOWCOLORSPACE(’Lab’) display a fully colored spectral locus in terms of CIE
L*a*b* coordinates.
showcolorspace(which, p) as above but plot the points whose xy- or a*b*-chromaticity
is given by the columns of p.
[IM,AX,AY] = showcolorspace(...) as above returns the spectral locus as an image
IM, with corresponding x- and y-axis coordinates AX and AY respectively.
Notes
• The colors shown within the locus only approximate the true colors, due to the
gamut of the display device.
See also
rg_addticks

191 Copyright Peter
c Corke 2017
showpixels
Show low resolution image
Displays a low resolution image in detail as a grid with colored lines between pixels
and numeric display of pixel values at each pixel. Useful for illustrating principles in
teaching.
Options
‘fmt’, F Format string (defaults to %d or %.2f depending on image type)

‘label’ Display axis labels (default true)
‘color’, C Text color (default ‘b’)
‘fontsize’, S Font size (default 12)
‘pixval’ Display pixel numeric values (default true)
‘tick’ Display axis tick marks (default true)
‘cscale’, C Color map scaling [min max] (defaults [0 1] or [0 255])
‘uv’, UV UV={u,v} vectors of u and v coordinates
‘infcolor’ show Inf values as red
‘nancolor’ show NaN values as red
‘hideinf’ don’t display value if Inf
‘hidenan’ don’t display value if Nan
‘contrast’ display text as white against dark squares
Notes
• This is meant for small images, say 10 × 10 pixels.
SiftPointFeature
SIFT point corner feature object
A subclass of OrientedScalePointFeature for SIFT features.
Methods


192 Copyright Peter
c Corke 2017

Properties
scale feature scale
image_id index of image containing feature
Properties of a vector of SiftCornerFeature objects are returned as a vector. If F is a

vector (N × 1) of SiftCornerFeature objects then F.u is a 2 × N matrix with each column
the corresponding u coordinate.
Notes
• SiftCornerFeature is a reference object.

• SiftCornerFeature objects can be used in vectors and arrays
• The SIFT algorithm is patented and not distributed with this toolbox. You can
download a SIFT implementation which this class can utilize. See README.SIFT.
References
“Distinctive image features from scale-invariant keypoints”, D.Lowe, Int. Journal on

Computer Vision, vol.60, pp.91-110, Nov. 2004.
See also
isift, PointFeature, ScalePointFeature, OrientedScalePointFeature, SurfPointFeature
SiftPointFeature.SiftPointFeature
Create a SIFT point feature object
f = SiftPointFeature() is a point feature object with null parameters.

193 Copyright Peter
c Corke 2017
f = SiftPointFeature(u, v) is a point feature object with specified coordinates.

f = SiftPointFeature(u, v, strength) as above but with specified strength.
f = SiftPointFeature(u, v, strength, scale) as above but with specified feature scale.
f = SiftPointFeature(u, v, strength, scale, theta) as above but with specified feature
orientation.
See also
isift
SiftPointFeature.match
Match SIFT point features

matches between the two vectors of SIFT features F and f2. Correspondence is based
on descriptor similarity.
SiftPointFeature.support
Support region of feature
out = F.support(im, w) is an image of the support region of the feature F, extracted

from the image im in which the feature appears. The support region is scaled to w × w
and rotated so that the feature’s orientation axis is upward.
out = F.support(images, w) as above but if the features were extracted from an image
sequence images then the feature is extracted from the appropriate image in the same
sequence.
[out,T] = F.support(images, w) as above but returns the pose of the feature as a 3 × 3
homogeneous transform in SE(2) that comprises the feature position and orientation.
F.support(im, w) as above but the support region is displayed.
See also
SiftPointFeature

194 Copyright Peter
c Corke 2017
SphericalCamera
Spherical camera class
A concrete class a spherical-projection camera.
Methods
project project world points




rho intrinsic: pixel dimensions (2 × 1) in metres


195 Copyright Peter
c Corke 2017
Note
• SphericalCamera is a reference object.

• SphericalCamera objects can be used in vectors and arrays
See also
Camera
SphericalCamera.SphericalCamera
Create spherical projection camera object
C = SphericalCamera() creates a spherical projection camera with canonic parame-

ters: f=1 and name=’canonic’.
C = CentralCamera(options) as above but with specified parameters.
Options

‘pixel’, S Pixel size: S × S or S(1)xS(2)
See also
Camera, CentralCamera, FisheyeCamera, CatadioptricCamera
SphericalCamera.plot_camera
Display camera icon in world view
C.plot_camera(T) draws the spherical image plane (unit sphere) at pose given by the
SE3 object T.
C.plot_camera(T, p) as above but also display world points, given by the columns of
p (3 × N), as small spheres.

196 Copyright Peter
c Corke 2017
Reference
“Spherical image-based visual servo and structure estimation”, p. I. Corke, in Proc.

IEEE Int. Conf. Robotics and Automation, (Anchorage), pp. 5550-5555, May 3-7
2010.
See also
SphericalCamera.project
pt = C.project(p, options) are the image plane coordinates for the world points p.
The columns of p (3 × N) are the world points and the columns of pt (2 × N) are the
corresponding spherical projection points, each column is phi (longitude) and theta
(colatitude).
Options
pose C.T.
See also
SphericalCamera.plot
SphericalCamera.sph
Implement spherical IBVS for point features
results = sph(T) results = sph(T, params)

Simulate IBVS with for a square target comprising 4 points is placed in the world XY
plane. The camera/robot is initially at pose T and is driven to the orgin.
Two windows are shown and animated:

197 Copyright Peter
c Corke 2017
1. The camera view, showing the desired view (*) and the
current view (o)
2. The external view, showing the target points and the camera
The results structure contains time-history information about the image plane, cam-
era pose, error, Jacobian condition number, error norm, image plane size and desired
feature locations.
The params structure can be used to override simulation defaults by providing ele-
ments, defaults in parentheses:
target_size - the side length of the target in world units (0.5)
target_center - center of the target in world coords (0,0,2)
niter - the number of iterations to run the simulation (500)

eterm - a stopping criteria on feature error norm (0)
lambda - gain, can be scalar or diagonal 6 × 6 matrix (0.01)
ci - camera intrinsic structure (camparam)
depth - depth of points to use for Jacobian, scalar for
all points, of 4-vector. If null take actual value

from simulation ([])
SEE ALSO: ibvsplot
SphericalCamera.sph2
Implement spherical IBVS for point features
results = sph(T) results = sph(T, params)

Simulate IBVS with for a square target comprising 4 points is placed in the world XY
plane. The camera/robot is initially at pose T and is driven to the orgin.
Two windows are shown and animated:
1. The camera view, showing the desired view (*) and the
current view (o)
2. The external view, showing the target points and the camera
The results structure contains time-history information about the image plane, cam-
era pose, error, Jacobian condition number, error norm, image plane size and desired
feature locations.
The params structure can be used to override simulation defaults by providing ele-
ments, defaults in parentheses:
target_size - the side length of the target in world units (0.5)

198 Copyright Peter
c Corke 2017
target_center - center of the target in world coords (0,0,3)
niter - the number of iterations to run the simulation (500)

eterm - a stopping criteria on feature error norm (0)
lambda - gain, can be scalar or diagonal 6 × 6 matrix (0.01)
ci - camera intrinsic structure (camparam)
depth - depth of points to use for Jacobian, scalar for
all points, of 4-vector. If null take actual value

from simulation ([])
SEE ALSO: ibvsplot
SphericalCamera.visjac_p
J = C.visjac_p(pt, z) is the image Jacobian (2N × 6) for the image plane points pt
(2 × N) described by phi (longitude) and theta (colatitude). The depth of the points
from the camera is given by z which is a scalar, for all points, or a vector (N × 1) for
each point.
The Jacobian gives the image-plane velocity in terms of camera spatial velocity.
Reference
“Spherical image-based visual servo and structure estimation”, P. I. Corke, in Proc.

IEEE Int. Conf. Robotics and Automation, (Anchorage), pp. 5550-5555, May 3-7
2010.
See also
ssd
Sum of squared differences
m = ssd(i1, i2) is the sum of squared differences between the two equally sized image
patches i1 and i2. The result m is a scalar that indicates image similarity, a value of
0 indicates identical pixel patterns and is increasingly positive as image dissimilarity
increases.

199 Copyright Peter
c Corke 2017
See also
zsdd, sad, ncc, isimilarity
stdisp
Display stereo pair
stdisp(L, R) displays the stereo image pair L and R in adjacent windows.

Two cross-hairs are created. Clicking a point in the left image positions black cross
hair at the same pixel coordinate in the right image. Clicking the corresponding world
point in the right image sets the green crosshair and displays the disparity [pixels].
See also
idisp, istereo
SurfPointFeature
SURF point corner feature object
A subclass of OrientedScalePointFeature for SURF features.
Methods


200 Copyright Peter
c Corke 2017
Properties
scale feature scale
image_id index of image containing feature
Properties of a vector of SurfCornerFeature objects are returned as a vector. If F is

a vector (N × 1) of SurfCornerFeature objects then F.u is a 2 × N matrix with each
column the corresponding u coordinate.
Notes
• SurfCornerFeature is a reference object.

• SurfCornerFeature objects can be used in vectors and arrays
Reference
“SURF: Speeded Up Robust Features”, Herbert Bay, Andreas Ess, Tinne Tuytelaars,
Luc Van Gool, Computer Vision and Image Understanding (CVIU), Vol. 110, No. 3,
pp. 346–359, 2008
See also
isurf, PointFeature, ScalePointFeature, OrientedScalePointFeature, SiftPointFeature
SurfPointFeature.SurfPointFeature
Create a SURF point feature object
f = SurfPointFeature() is a point feature object with null parameters.

f = SurfPointFeature(u, v) is a point feature object with specified coordinates.
f = SurfPointFeature(u, v, strength) as above but with specified strength.
f = SurfScalePointFeature(u, v, strength, scale) as above but with specified feature
scale.
f = SurfPointFeature(u, v, strength, scale, theta) as above but with specified feature
orientation.

201 Copyright Peter
c Corke 2017
See also
isurf, OrientedScalePointFeature
SurfPointFeature.match
Match SURF point features

matches between the two vectors of SURF features F and f2. Correspondence is based
on descriptor similarity.
[m,C] = F.match(f2, options) as above but returns a correspodence matrix where each
row contains the indices of corresponding features in F and f2 respectively.
Options
‘thresh’, T match threshold

‘top’, N Take strongest N matches
Notes
• to obtain all matches use ‘top’, Inf
See also
FeatureMatch
SurfPointFeature.support
Support region of feature
out = F.support(im, w) is an image of the support region of the feature F, extracted

from the image im in which the feature appears. The support region is scaled to w × w
and rotated so that the feature’s orientation axis is upward.
out = F.support(images, w) as above but if the features were extracted from an image
sequence images then the feature is extracted from the appropriate image in the same
sequence.
[out,T] = F.support(images, w) as above but returns the pose of the feature as a 3 × 3
homogeneous transform in SE(2) that comprises the feature position and orientation.
F.support(im, w) as above but the support region is displayed.

202 Copyright Peter
c Corke 2017
See also
SurfPointFeature
tb_optparse
Standard option parser for Toolbox functions
optout = tb_optparse(opt, arglist) is a generalized option parser for Toolbox func-

tions. opt is a structure that contains the names and default values for the options, and
arglist is a cell array containing option parameters, typically it comes from VARAR-
GIN. It supports options that have an assigned value, boolean or enumeration types
(string or int).
The software pattern is:
function(a, b, c, varargin)
opt.foo = false;
opt.bar = true;
opt.blah = [];
opt.stuff = {};
opt.choose = {’this’, ’that’, ’other’};
opt.select = {’#no’, ’#yes’};
opt = tb_optparse(opt, varargin);
Optional arguments to the function behave as follows:
‘foo’ sets opt.foo := true

‘nobar’ sets opt.foo := false
‘blah’, 3 sets opt.blah := 3
‘blah’, {x,y} sets opt.blah := {x,y}
‘that’ sets opt.choose := ‘that’
‘yes’ sets opt.select := (the second element)
‘stuff’, 5 sets opt.stuff to {5}
‘stuff’, {’k’,3} sets opt.stuff to {’k’,3}
and can be given in any combination.

If neither of ‘this’, ‘that’ or ‘other’ are specified then opt.choose := ‘this’. Alternatively
if:
opt.choose = {[], ’this’, ’that’, ’other’};
then if neither of ‘this’, ‘that’ or ‘other’ are specified then opt.choose := []

If neither of ‘no’ or ‘yes’ are specified then opt.select := 1.
Note:
• That the enumerator names must be distinct from the field names.

203 Copyright Peter
c Corke 2017
• That only one value can be assigned to a field, if multiple values are required
they must placed in a cell array.
• To match an option that starts with a digit, prefix it with ‘d_’, so the field ‘d_3d’
matches the option ‘3d’.
• opt can be an object, rather than a structure, in which case the passed options are
assigned to properties.
The return structure is automatically populated with fields: verbose and debug. The
following options are automatically parsed:
‘verbose’ sets opt.verbose := true

‘verbose=2’ sets opt.verbose := 2 (very verbose)
‘verbose=3’ sets opt.verbose := 3 (extremeley verbose)
‘verbose=4’ sets opt.verbose := 4 (ridiculously verbose)
‘debug’, N sets opt.debug := N
‘showopt’ displays opt and arglist
‘setopt’, S sets opt := S, if S.foo=4, and opt.foo is present, then opt.foo is set to 4.
The allowable options are specified by the names of the fields in the structure opt. By
default if an option is given that is not a field of opt an error is declared.
[optout,args] = tb_optparse(opt, arglist) as above but returns all the unassigned op-
tions, those that don’t match anything in opt, as a cell array of all unassigned arguments
in the order given in arglist.
[optout,args,ls] = tb_optparse(opt, arglist) as above but if any unmatched option
looks like a MATLAB LineSpec (eg. ‘r:’) it is placed in ls rather than in args.
[objout,args,ls] = tb_optparse(opt, arglist, obj) as above but properties of obj with
matching names in opt are set.
testpattern
Create test images
im = testpattern(type, w, args) creates a test pattern image. If w is a scalar the image

is w × w else w(2)xW(1). The image is specified by the string type and one or two
(type specific) arguments:
‘rampx’ intensity ramp from 0 to 1 in the x-direction. args is the number of cycles.
‘rampy’ intensity ramp from 0 to 1 in the y-direction. args is the number of cycles.
‘sinx’ sinusoidal intensity pattern (from -1 to 1) in the x-direction. args is the number of
cycles.
‘siny’ sinusoidal intensity pattern (from -1 to 1) in the y-direction. args is the number of
cycles.

204 Copyright Peter
c Corke 2017
‘dots’ binary dot pattern. args are dot pitch (distance between centres); dot diameter.
‘squares’ binary square pattern. args are pitch (distance between centres); square side length.
‘line’ a line. args are theta (rad), intercept.
Examples
A 256 × 256 image with 2 cycles of a horizontal sawtooth intensity ramp:

testpattern(’rampx’, 256, 2);
A 256 × 256 image with a grid of dots on 50 pixel centres and 20 pixels in diameter:
testpattern(’dots’, 256, 50, 25);
Notes
• With no output argument the testpattern in displayed using idisp.
See also
idisp
Tracker
Track points in image sequence
This class assigns each new feature a unique identifier and tracks it from frame to frame
until it is lost. A complete history of all tracks is maintained.
Methods
plot Plot all tracks

tracklengths Length of all tracks
Properties
track A vector of structures, one per active track.

history A vector of track history structures with elements id and uv which is the path of the
feature.

205 Copyright Peter
c Corke 2017
See also
PointFeature
Tracker.Tracker
Create new Tracker object
T = Tracker(im, C, options) is a new tracker object. im (H × W × S) is an image

sequence and C (S × 1) is a cell array of vectors of PointFeature subclass objects. The
elements of the cell array are the point features for the corresponding element of the
image sequence.
During operation the image sequence is animated and the point features are overlaid
along with annotation giving the unique identifier of the track.
Options
‘radius’, R Search radius for feature in next frame (default 20)

‘nslots’, N Maximum number of tracks (default 800)
‘thresh’, T Similarity threshold (default 0.8)
‘movie’, M Write the frames as images into the folder M as with sequential filenames.
Notes
• The ‘movie’ options saves frames as files NNNN.png.

• When using ‘movie’ option ensure that the window is fully visible.
• To convert frames to a movie use a command like:
ffmpeg -r 10 -i %04d.png out.avi
See also
PointFeature
Tracker.char
Convert to string
s = T.char() is a compact string representation of the Tracker parameters and status.

206 Copyright Peter
c Corke 2017
Tracker.display
Display value
T.display() displays a compact human-readable string representation of the Tracker

object
Notes
expression is a Tracker object and the command has no trailing semicolon.
See also
Tracker.char
Tracker.plot
Show feature trajectories
T.plot() overlays the tracks of all features on the current plot.
Tracker.tracklengths
Length of all tracks
T.tracklengths() is a vector containing the length of every track.
tristim2cc
Tristimulus to chromaticity coordinates
cc = tristim2cc(tri) is the chromaticity coordinate (1 × 2) corresponding to the tris-

timulus tri (1 × 3). If tri is RGB then cc is rg, if tri is XYZ then cc is xy. Multiple
tristimulus values can be given as rows of tri (N × 3) in which case the chromaticity
coordinates are the corresponding rows of cc (N × 2).

207 Copyright Peter
c Corke 2017
[c1,C2] = tristim2cc(tri) as above but the chromaticity coordinates are returned in

separate vectors, each N × 1.
out = tristim2cc(im) is the chromaticity coordinates corresponding to every pixel in
the tristimulus image im (H ×W × 3). out (H ×W × 2) has planes corresponding to r
and g, or x and y.
[o1,o2] = tristim2cc(im) as above but the chromaticity is returned as separate images
(H ×W ).
upq
Central image moments
m = upq(im, p, q) is the PQth central moment of the image im. That is, the sum of
I(x,y).(x-x0)p .(y-y0)q where (x0,y0) is the centroid.
Notes
• The central moments are invariant to translation.
See also
upq_poly, mpq, npq
upq_poly
Central polygon moments
m = upq_poly(v, p, q) is the PQth central moment of the polygon with vertices de-
scribed by the columns of v.
Notes
clockwise).
still be correct.

208 Copyright Peter
c Corke 2017
• If the first and last point in the list are the same, they are considered as a single
vertex.
• The central moments are invariant to translation.
See also
upq, mpq_poly, npq_poly
usefig
figure windows
usefig(’Foo’) makes figure ‘Foo’ the current figure, if it doesn’t exist create it.
h = usefig(’Foo’) as above, but returns the figure handle
VideoCamera
Abstract class to read from local video camera
A concrete subclass of ImageSource that acquires images from a local camera using the
MATLAB Image Acquisition Toolbox (imaq). This Toolbox provides a multiplatform
interface to a range of cameras, and this class provides a simple wrapper.
This class is not intended to be used directly, instead use the factory method Video
which will return an instance of this class if the Image Acquisition Toolbox is installed,
for example
vid = VideoCamera();
Methods

size Size of image

209 Copyright Peter
c Corke 2017
See also
VideoCamera, ImageSource, AxisWebCamera, Movie
VideoCamera_fg
Class to read from local video camera
A concrete subclass of ImageSource that acquires images from a local camera using a
simple open-source frame grabber interface.
This class is not intended to be used directly, instead use the factory method Video-
Camera.which will return an instance of this class if the interface is supported on your
platform (Mac or Linux), for example
vid = VideoCamera.amera();
Methods

size Size of image
See also
ImageSource, AxisWebCamera, Movie
VideoCamera_fg.VideoCamera_fg
Video camera constructor
V = VideoCamera_fg.CAMERA, OPTIONS) is a VideoCamera_fg.object that ac-

quires images from the local video camera specified by the string CAMERA.
If CAMERA is ‘?’ a list of available cameras, and their characteristics is displayed.
Options

210 Copyright Peter
c Corke 2017

‘id’, I ID of camera
Notes:
VideoCamera_fg.char
Convert to string
V.char() is a string representing the state of the camera object in human readable form.
VideoCamera_fg.close
V.close() closes the connection to the camera.
VideoCamera_fg.grab
im = V.grab() acquires an image from the camera.
Notes
• the function will block until the next frame is acquired.

211 Copyright Peter
c Corke 2017
VideoCamera_IAT
Class to read from local video camera
A concrete subclass of ImageSource that acquires images from a local camera using the
MATLAB Image Acquisition Toolbox (imaq). This Toolbox provides a multiplatform
interface to a range of cameras, and this class provides a simple wrapper.
This class is not intended to be used directly, instead use the factory method Video
which will return an instance of this class if the Image Acquisition Toolbox is installed,
for example
vid = VideoCamera();
Methods

size Size of image
See also
VideoCamera, ImageSource, AxisWebCamera, Movie
VideoCamera_IAT.VideoCamera_IAT
Video camera constructor
v = Video_IAT(camera, options) is a Video object that acquires images from the local
video camera specified by the string camera.
Options

‘id’, I ID of camera

212 Copyright Peter
c Corke 2017
Notes:
VideoCamera_IAT.char
Convert to string
V.char() is a string representing the state of the camera object in human readable form.
VideoCamera_IAT.close
V.close() closes the connection to the camera.
VideoCamera_IAT.grab
im = V.grab() acquires an image from the camera.
Notes
• the function will block until the next frame is acquired.
VideoCamera_IAT.list
available adaptors and cameras
VideoCamera_IAT.preview
Control image preview
V.preview(true) enables camera preview in a separate window

213 Copyright Peter
c Corke 2017
xaxis
Set X-axis scaling
xaxis(max) set x-axis scaling from 0 to max.

xaxis(min, max) set x-axis scaling from min to max.
xaxis([min max]) as above.
xaxis restore automatic scaling for x-axis.
See also
yaxis
xyzlabel
Label X, Y and Z axes
XYZLABEL label the x-, y- and z-axes with ‘X’, ‘Y’, and ‘Z’ respectiveley
yaxis
Y-axis scaling
yaxis(max) set y-axis scaling from 0 to max.

yaxis(min, max) set y-axis scaling from min to max.
yaxis([min max]) as above.
yaxis restore automatic scaling for y-axis.
See also
yaxis

214 Copyright Peter
c Corke 2017
YUV
Class to read YUV4MPEG file
A concrete subclass of ImageSource that returns images from a YUV4MPEG format

uncompressed video file.
Methods

size Size of image
Properties
curFrame The index of the frame just read
See also
ImageSource, Video
SEE ALSO: Video
YUV.YUV
YUV4MPEG sequence constructor
y = YUV(file, options) is a YUV4MPEG object that returns frames from the yuv4mpeg
format file file. This file contains uncompressed color images in 4:2:0 format, with a
full resolution luminance plane followed by U and V planes at half resolution both
directions.
Options

‘scale’, S Subsample the image by S in both directions
‘skip’, S Read every Sth frame from the movie

215 Copyright Peter
c Corke 2017
YUV.char
Convert to string
M.char() is a string representing the state of the movie object in human readable form.
YUV.close
M.close() closes the connection to the movie.
YUV.grab
Acquire next frame from movie
im = Y.grab(options) is the next frame from the file.

[y,u,v] = y.grab(options) is the next frame from the file
Options
‘skip’, S Skip frames, and return current+S frame (default 1)

‘rgb’ Return as an RGB image, y image is downsized by two (default).
‘rgb2’ Return as an RGB image, u and v images are upsized by two.
‘yuv’ Return y, u and v images.
Notes
• If no output argument given the image is displayed using IDISP.
• For the ‘yuv’ option three output arguments must be given.

216 Copyright Peter
c Corke 2017
yuv2rgb
YUV format to RGB
[r,g,b] = yuvread(y, u, v) rgb = yuvread(y, u, v)

Returns the equivalent RGB image from YUV components. The Y image is halved in
resolution.
yuv2rgb2
YUV format to RGB
[r,g,b] = yuvread2(y, u, v) rgb = yuvread(y, u, v)

Returns the equivalent RGB image from YUV components. The UV images are dou-
bled in resolution so the resulting color image is original size.
zcross
Zero-crossing detector
iz = zcross(im) is a binary image with pixels set where the corresponding pixels in the
signed image im have a zero crossing, a positive pixel adjacent to a negative pixel.
Notes
• Can be used in association with a Lapalacian of Gaussian image to determine

edges.
See also
ilog

217 Copyright Peter
c Corke 2017
zncc
Normalized cross correlation
m = zncc(i1, i2) is the zero-mean normalized cross-correlation between the two equally
sized image patches i1 and i2. The result m is a scalar in the interval -1 to 1 that
indicates similarity. A value of 1 indicates identical pixel patterns.
Notes
• The zncc similarity measure is invariant to affine changes in image intensity

(brightness offset and scale).
See also
ncc, sad, ssd, isimilarity
zsad
Sum of absolute differences
m = zsad(i1, i2) is the zero-mean sum of absolute differences between the two equally
sized image patches i1 and i2. The result m is a scalar that indicates image similarity,
a value of 0 indicates identical pixel patterns and is increasingly positive as image
dissimilarity increases.
Notes
• The zsad similarity measure is invariant to changes in image brightness offset.
See also
sad, ssd, ncc, isimilarity

218 Copyright Peter
c Corke 2017
zssd
Sum of squared differences
m = zssd(i1, i2) is the zero-mean sum of squared differences between the two equally
sized image patches i1 and i2. The result m is a scalar that indicates image similarity,
a value of 0 indicates identical pixel patterns and is increasingly positive as image
dissimilarity increases.
Notes
• The zssd similarity measure is invariant to changes in image brightness offset.
See also
sdd, sad, ncc, isimilarity

219 Copyright Peter
c Corke 2017

Vision

Uploaded by

Copyright:

Available Formats

Vision

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vision

Uploaded by

Copyright:

Available Formats

Machine Vision

This, the fourth major release of the Toolbox, repre-

Machine Vision Toolbox 4.1 for MATLAB3 Copyright Peter

2 Functions and classes 12

Machine Vision Toolbox 4.1 for MATLAB4 Copyright Peter

Machine Vision Toolbox 4.1 for MATLAB5 Copyright Peter

Machine Vision Toolbox 4.1 for MATLAB6 Copyright Peter

Machine Vision Toolbox 4.1 for MATLAB7 Copyright Peter

Machine Vision Toolbox 4.1 for MATLAB8 Copyright Peter

Color Image sources

Camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Type changing

Machine Vision Toolbox 4.1 for MATLAB9 Copyright Peter

Spatial operators zncc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

Non-linear operators Line features

Machine Vision Toolbox 4.1 for MATLAB10 Copyright Peter

Multiview Image generation

Machine Vision Toolbox 4.1 for MATLAB11 Copyright Peter

Machine Vision Toolbox 4.1 for MATLAB12 Copyright Peter

1.1 Changes in MVTB 4

MVTB 4 is largely backward compatible with MVTB 3.

1.1.1 Incompatible changes

• xycolorspace replaced with showcolorspace(’xy’). It can also dis-

• iconv changed to iconvolve

• RegionFeature property shape has been renamed to aspect

1.1.2 New features

• Bundle adjustment is supported by the new BundleAdjust class.

Machine Vision Toolbox 4.1 for MATLAB13 Copyright Peter

• cmf2xyz for spectral input multiplies by ∆λ

1.2 How to obtain the Toolbox

1.2.1 From .mltbx file

1.2.2 From .zip file

Machine Vision Toolbox 4.1 for MATLAB14 Copyright Peter

• contrib.zip A small number of Toolbox functions depend on third party

• images2.zip This is a large file (150MB) containing the mosaic, campus,

1.2.3 MATLAB OnlineTM

MATLAB Online does not support MEX files.

Machine Vision Toolbox 4.1 for MATLAB15 Copyright Peter

This document vision.pdf is a comprehensive manual that describes all functions

1.3 Compatible MATLAB versions

1.4 Use in teaching

1.5 Use in research

Machine Vision Toolbox 4.1 for MATLAB16 Copyright Peter

1.7 Related software

1.7.1 Image Processing Toolbox

1.7.2 Computer Vision System Toolbox

Machine Vision Toolbox 4.1 for MATLAB17 Copyright Peter

GNU Octave (www.octave.org) is an impressive piece of free software that implements

1.7.4 Robotics Toolbox

Machine Vision Toolbox 4.1 for MATLAB18 Copyright Peter

Functions and classes

Machine Vision Toolbox 4.1 for MATLAB19 Copyright Peter

the left image is red, and the right image is cyan.

Load the left and right images

then display the anaglyph for viewing with red-cyan glasses

• Robotics, Vision & Control, Section 14.3, P. Corke, Springer 2011.

Machine Vision Toolbox 4.1 for MATLAB20 Copyright Peter

Elements of the structure are:

.id decoded id of the tag in the range 1-255

• implementation is a mex file

• apriltags is open-source software from University of Michigan