Scribd
Upload
61 views57 pages

EECS 556 - Image Processing - W 09 Interpolation: - Interpolation Techniques - B Splines

This document discusses image interpolation techniques. It introduces concepts like interpolation, resampling, and image warping. It covers various interpolation methods like nearest neighbor, bilinear, and b-spline interpolation. B-spline interpolation is described as a popular choice as it produces interpolants that are n-order differentiable and have short spatial support, minimizing computation.

Uploaded by

sanjapatil
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
61 views57 pages

EECS 556 - Image Processing - W 09 Interpolation: - Interpolation Techniques - B Splines

This document discusses image interpolation techniques. It introduces concepts like interpolation, resampling, and image warping. It covers various interpolation methods like nearest neighbor, bilinear, and b-spline interpolation. B-spline interpolation is described as a popular choice as it produces interpolants that are n-order differentiable and have short spatial support, minimizing computation.

Uploaded by

sanjapatil
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 57

EECS556 ImageProcessing W09

Interpolation

Interpolationtechniques
Bsplines

Whatisimageprocessing?
Imageprocessingistheapplicationof2Dsignalprocessing
methodstoimages

Imagerepresentationandmodeling
Imageenhancement
Imagerestoration/imagefiltering
Compression

(computervision)

Imageenhancement
Accentuatecertaindesiredfeaturesfor
subsequentanalysisordisplay.

Contraststretching/dynamicrangeadjustment
Colorhistogramnormalization
Noisereduction
Sharpening
Edgedetection
Cornerdetection
Imageinterpolation

Whatisinterpolation?
givenadiscretespaceimagegd[n,m]
Thiscorrespondstosamplesofsomecontinuous
spaceimagega(x,y)
ComputevaluesoftheCSimagega(x,y)at(x,y)
locationsotherthanthesamplelocations.
ga(x,y)

gd[n,m]

gd[n,m]

Whatisinterpolation?
Interpolationiscriticalfor:
Displaying
Imagezooming
Warping
Coding
Motionestimation

Isperfectrecoveryalwayspossible?
No,(neveractually),unlesssomeassumptions
onga(x,y)aremade
Thereareuncountableinfinitecollectionof
functionsga(x,y)thatagreewithgd[n,m]at
thesamplepoints
Whataretheseassumptions?
ga mustbebandlimited
SamplingratemustsatisfyNyquist theorem

ga mustbebandlimited
Thereexists

suchthat

SamplingratemustsatisfyNyquist
theorem

IdealUniformRectilinearSampling
2Duniformlysampledgrid

IdealUniformRectilinearSampling
ga(x,y)

gs(x,y)

Howtorecoverga ?

ga(x,y,)

Howtorecoverga ?

Sinc interpolation
Wecanrecoverga(x,y)byinterpolatingthe
samplesgd[n,m]usingsinc functions

Interpolation
Sampledgs

MATLABsinterp2

Whatstheproblemwithsinc
interpolation?
Realworldimagesneednotbeexactlyband
limited.
Unboundedsupport
Summationsrequireinfinitelymanysamples
Computationallyveryexpensive

Linearinterpolation
sinc interpolationformula

Linearinterpolation:

h(x,y)=interpolationkernel
Whythisisalinearinterpolation?

Linearfunctionofthesamplesgd[n,m]

Linearinterpolation
Consideralternativelinearinterpolation
schemesthataddressissueswith:
Unboundedsupport
Summationsrequireinfinitelymanysamples
Computationallyveryexpensive

Basicpolynomialinterpolation
Herekernelsthatarepiecewisepolynomials
Zeroorderornearestneighborinterpolation
Usethevalueatthenearestsamplelocation
Kernelsarezeroorderpolynomials

Nearestneighborinterpolation

Linearorbilinearinterpolation
tri functionisapiecewiselinearfunctionofits
(spatial)arguments

/2

/2

Linearorbilinearinterpolation
Foranypoint(x,y),wefindthefournearestsamplepoint
fitapolynomialoftheform
Estimatealphasfromsystemof4equationsin4unknowns:

Interpolatingformula(x [0,1] [0,1]):

Linearorbilinearinterpolation

Ifx [1,1] [1,1]:

Linearorbilinearinterpolation

LinearinterpolationbyDelaunay
triangulation
Usealinearfunctionwithin
eachtriangle(threepoints
determineaplane).

MATLABsgriddata with
thelinearopt

Whatstheproblemzeroorderor
bilinearinterpolation?
neitherthezeroorderorbilinearinterpolator
aredifferentiable.
Theyarenotcontinuousandsmooth

Desirablepropertiesofinterpolators
selfconsistency:
Continuousandsmooth(differentiable):

Shortspatialextenttominimizecomputation

Desirablepropertiesofinterpolators

Frequencyresponseapproximatelyrect().
symmetric
Shiftinvariant
Minimumsidelobes toavoidringingartifacts

Continuousandsmooth
Largesidelobes

Nonsmooth
Smallsidelobes

Nonsmooth;nonshiftinvariant

Polynomialinterpolation
Cubic
interpolationkernel:

Quadratic
Bsplines

Bspline interpolation
Motivation:
norderdifferentiable
Shortspatialextenttominimizecomputation

Background
Splines arepiecewisepolynomialswithpiecesthat
aresmoothlyconnectedtogether
Thejoiningpointsofthepolynomialsarecalledknots

Background
Foraspline ofdegreen,eachsegmentisapolynomial
ofdegreen
Doi needn+1coefficienttodescribeeachpiece?

Additionalsmoothnessconstraintimposesthecontinuityof
thespline anditsderivativesuptoorder(n1)attheknots
onlyonedegreeoffreedompersegment!

Bspline expansion
Splines areuniquelycharacterizedintermsof
aBspline expansion

integershiftsofthecentralBspline ofdegreen
=centralBspline =basis(B)spline

Bsplines
Symmetrical
bellshapedfunctions
constructedfromthe(n+1)foldconvolution

Bsplines

Bsplines
norderBspline:

Derivedusing:

Bspline
UsingcubicBsplines isapopularchoice:

Important:compactlysupported!

CubicBspline anditsunderlyingcomponents

support

Bspline expansion
Eachspline isunambiguously
characterizedbyitssequenceof
Bspline coefficientsc(k)

PropertiesofBspline expansion
Discretesignalrepresentation!(eventhoughthe
underlyingmodelisacontinuousrepresentation).

Easytomanipulate;
E.g.,derivatives:

Bspline interpolation
Sofar:Bspline modelofagiveninputsignal
s(x)
sd[n]
Interpolationproblem:findcoefficientsc(k)
suchthatspline functiongoesthroughthe
datapointsexactly
Thatis,reconstructsignalusingaspline representation!

Reconstructsignalusingaspline
representation
sd [n]

Bspline interpolation

Relationshipbetweencoefficientsandsamples

sd [n]
=discreteBspline kernel
ObtainedbysamplingtheBspline ofdegreenexpandedbya
factorofm(m=1)

CardinalBspline interpolation

Cardinalsplines

2Dsplines inimages
tensorproductbasisfunctions

Bspline interpolation
Conclusion:
norderdifferentiable
Shortspatialextenttominimize
computation

Generalimagespatialtransformations
transformthespatialcoordinatesofanimagef(x,y)
sothat(afterbeingtransformed),itbettermatches
anotherimage
Ex:warpingbrainimagesintoastandardcoordinate
systemtofacilitateautomaticimageanalysis

imagespatialtransformations

Generalimagespatialtransformations
Formoftransformations:
Shift only g(x,y)=f(xx0,yy0)
Affineg(x,y)=f(ax+by,cx +dy)
thinplatesplines (describessmoothwarpings using
controlpoints)
Generalspatialtransformation(e.g.,morphingor
warpinganimage)

MATLABsinterp2orgriddata functions

Interpolationiscritical!

You might also like