CT036-3-3-IPPR: Image Processing, Computer Vision

and Pattern Recognition (VE1)

Week 11 & 12
Representation and Description
 Motivation

Task: Describe the region based on the chosen representation

Image acquisition => digital image

↓
Preprocessing => better image
↓
Segmentation => basic features
↓
Representation and description => advanced features
↓
Object recognition (classification, clustering)

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 2
 Topic & Structure of the lesson

• Representation schemes

• Boundary descriptors

• Regional descriptors

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 3
CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 4
CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 5
CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 6
 Representation – Chain Codes

Chain Code
– represent a boundary by a connected sequence of straight-line
segment of specified length and direction
– based on 4-or- 8 connectivity of the segments
– can be generated following a boundary in a clockwise direction
and assigning a direction to the segments connecting every pair of
pixels

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 7
 Representation –
Resampling for Chain Codes

Spacing of the
sampling grid affect
the accuracy of the
representation

0033…01 0766…12
(4-directional) (8-directional)

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Slide

Representation and Description SLIDE 8
 Image Representation

Normalization for Chain Codes

• With respect to starting point, normalization for starting position –
shape number
– Make the chain code a circular sequence
– Redefine the starting point which gives an integer of minimum
magnitude
– E.g. 101003333222 normalized to 003333222101

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 9
 Image Representation

Normalization for Chain Codes

• Normalization for rotation – first difference (FD)
– Counting (in counterclockwise direction) the number of
direction changes that separate two adjacent element
of the code
– E.g. FD of a 4-direction chain code 10103322 is
33133030
– The first difference of smallest magnitude, E.g.
33133030 normalized to 03033133

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 10
 Boundary descriptors

Boundary information can be used directly, or converted into features that describe properties of a region,
such as
• Length of the boundary
• Orientation of straight line joining its extreme points
• Number of concavities in the boundary
• Diameter:
– Lead to Major axis and minor axis (Perpendicular to major)
– Basic Rectangle
• Curvature
– the rate of change of slope (the difference between the slope of adjacent boundary segments)
• Shape Number:
-Smallest integers of first difference chain code

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 11
 Boundary descriptors –
Shape Number
Circular chain code

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 12
 Boundary descriptors –
Shape Number

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 13
 Representation –
Polygonal Approximation

➢ Approximate a digital boundary with arbitrary

accuracy by a polygon;

➢ Goal: to capture the “essence” of the boundary

shape with the fewest possible polygonal
segments.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Slide

Representation and Description SLIDE 14
 Representation –
Polygonal Approximation

Polygon Approximation:
– Dominant Points
– Splitting

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 15
 Representation –
Polygonal Approximation
Splitting Technique

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 16
 Representation –
The skeleton of region

• Skeletonization/Thinning:
– Medial Axis Transform (MAT)
• Point p is belong to medial (Region R and Border B):
– Has more than one closest neighbor in B

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 17
 Representation –
The skeleton of region

• Example of Thinning

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 18
 Regional descriptors

❑ Simple
❑ Topological
❑ Texture
❑ Statistical
❑ Spectral
❑ Moments

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 19
 Regional descriptors

Some simple Regional Descriptor:

❖ Area (Number of pixels)
❖ Perimeter (Length of boundary)
❖ Compactness (Perimeter2/Area)
❖ Gray level measures: mean, median, max etc.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 20
 Regional Descriptors –
Topology Description

Topology ➔The study of properties of a figure that are

unaffected by any deformation

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 21
 Regional descriptors –
Topology Descriptor

Number of holes (H):

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 22
 Regional descriptors –
Topology Descriptor
Number of connected components (C) :

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 23
 Regional descriptors –
Topology Descriptor

– Euler number (E=C-H) :

• Invariants to several operators.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 24
 Regional descriptors –
Topology Descriptor

•Topology Descriptor:
– Polygonal net:
• V: # of vertices (7)
• Q: # of edges (11)
• F: # of faces (2)
•E=C-H=V-Q+F
– C=1 , H=3

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 25
 Regional descriptors –
Topology Descriptor
Original Threshold

Large Skeleton
Connected
Region

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 26
 Regional descriptors – Texture

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 27
 Regional descriptors –
Statistical Texture
➢ Histogram based - 1st order grey level statistics
- Statistical moments
- Uniformity
- Average entropy
➢ Co-occurence matrix based - 2nd order grey level
statistics
- Maximum probability etc.

A drawback with using the histogram for texture analysis is

that no positional information is used. Only gives information
not if “patterned”.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 28
 Regional descriptors –
Statistical Texture
Histogram Based Descriptors
Statistical Moments ➔Texture feature from 1st order statistics

❖ Properties of the grey level histogram of an image or region, used

when calculating statistical moments.
➢ z - discrete random variable representing discrete grey levels in the
range [0, L − 1].
➢ p(zi) - normalised histogram component corresponding to the ith value
of z.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 29
 Regional descriptors –
Statistical Texture
Histogram Based Descriptors
Uniformity and Average Entropy
❖ Also use the histogram properties z and p(zi).

U is maximum for an image in which all grey levels are equal (maximally
uniform).

Measure of variability. Is 0 for constant images

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 30
 Regional descriptors –
Statistical Texture
▪ Co-occurence Matrix➔ 2nd order grey level statistics

✓ From GLCM (Grey Level Co-ocurrence Matrix)

• A pixel and its neighbor
✓ Needs:
• A Positioning Operator, P.
• GLCM(i,j): # of times that points with gray level

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 31
 Regional descriptors –
Statistical Texture
Co-occurence Matrix

• For an image with N grey levels, and P, a positional operator,

generate A, a N × N matrix, where ai,j is the number of times a pixel
with grey level value zi is in relative position P to grey level value zj.

• Divide all elements in A with the sum of all elements in A (the

number of times P was satisfied in the image). This gives a new
matrix C, where ci,j is the probability that a pair of pixels fulfilling P
has grey level values zi and zj, which is called the co-occurrence
matrix.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 32
 Regional descriptors –
Statistical Texture
▪ Co-occurence Matrix

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 33
 Regional descriptors –
Statistical Texture
Co-occurence Matrix Descriptors

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 34
 Regional descriptors –
Statistical Texture
Example Co-occurence Matrix Descriptors

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 35
 Regional descriptors –
Statistical Texture

Figure: Match sample image with correct co-occurrence matrix

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 36
 Relational Descriptors

❑ Primitive elements
❑ Directed line segments
❑ Tree

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 37
 Relational descriptors

Primitive elements:
 We introduced rewriting rules for describing texture.
 Here, we expand that concept in the context of
relational descriptors.
 Capture the basic repetitive patterns in a boundary
or region based on rewriting rules.

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 38
 Relational descriptors

❑ The staircase structure

❑ Formulate a recursive relationship involving these primitive elements
❑ Use the rewriting rules (1) S→aA; (2) A →bS and (3) A →b

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 39
 Relational descriptors

Directed line segments:

 Follow the contour of an object and code the results with
segments of specified direction and/or length
 String descriptions are best suited for applications in which
connectivity of primitives can be expressed in a head-to-tail

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 40
 Relational descriptors

• Tree descriptors
– Definition
• A unique root node.
• The remaining nodes are partitioned into m disjoined sets T1,…, T m
(subtree)
– The tree frontier: the set of nodes at the bottom of the tree (the leaves)

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 41
 Relational descriptors

Generally, two types of

information in a tree are
important:
◆ Information about a
node.
◆ Information relating a
node to its neighbors.

Relationship: inside of …

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 42
 Reference

• Gonzalez & Woods, Digital Image Processing

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 43
 Question and Answer Session

Q&A

CT036-3-3-IPPR: Image Processing, Computer Vision and Pattern Recognition Representation and Description SLIDE 44