Scribd
Upload
7 views64 pages

Representation and Description

The document discusses image features representation and description, emphasizing the importance of feature extraction for object classification. It covers natural and artificial features, boundary representation methods like chain code, and various algorithms for component labeling and regional descriptors. Additionally, it highlights the significance of shape, geometrical, and texture features in digital image processing.

Uploaded by

Karuna Gowda
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
7 views64 pages

Representation and Description

The document discusses image features representation and description, emphasizing the importance of feature extraction for object classification. It covers natural and artificial features, boundary representation methods like chain code, and various algorithms for component labeling and regional descriptors. Additionally, it highlights the significance of shape, geometrical, and texture features in digital image processing.

Uploaded by

Karuna Gowda
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/ 64

Digital Image Processing

S. Sridhar, Associate Professor,


Department of Information Science and
Technology, College of Engineering
Guindy Campus, Anna University,
Chennai

© Oxford University Press 2011


Chapter 10
Image Features Representation
and Description

© Oxford University Press 2011


What is a feature?
• Any characteristic or primitive of an object that
helps to distinguish or discriminate an object
from other objects is called an image feature.
• Natural features These are visual appearances
of the image that are natural to the object, such
as brightness and texture.
• Artificial features These are derived features
that are obtained using image manipulations.
Amplitude histograms and frequency spectrums
are examples of this category

© Oxford University Press 2011


Feature Extraction

• Feature extraction is a process of


extraction and generation of features to
assist the task of object classification. This
phase is critical because the quality of the
features influences the classification task.

© Oxford University Press 2011


© Oxford University Press 2011
Characteristics of a Good Feature

© Oxford University Press 2011


© Oxford University Press 2011
© Oxford University Press 2011
BOUNDARY REPRESENTATION

• Chain Code

© Oxford University Press 2011


Tracking Process

© Oxford University Press 2011


© Oxford University Press 2011
Advantages of Chain code

© Oxford University Press 2011


Disadvantage

© Oxford University Press 2011


DCC

© Oxford University Press 2011


Polygonal Approximation

© Oxford University Press 2011


Merging Method

© Oxford University Press 2011


Signatures

• A signature is a 1D functional
representation. It is a plot of the distance
from the centroid to the boundary as a
function of the angle.

© Oxford University Press 2011


© Oxford University Press 2011
Bending Energy

• Boundary curvature is the differential of the y-s


curve, i.e., the curvature at every point in the
boundary. The integration of all squared
curvature values along the entire contour gives a
single descriptor called bounding energy (or
bending energy). This is computed in the
discrete domain as the sum of squares of the
border curvature c(x) over the boundary of
length L.

© Oxford University Press 2011


© Oxford University Press 2011
Statistical Moments

© Oxford University Press 2011


Region Representation
• 1. Region decomposition
• 2. Bounding regions
• 3. Internal features

© Oxford University Press 2011


Thinning Algorithms

© Oxford University Press 2011


Contd…

© Oxford University Press 2011


BOUNDARY DESCRIPTIONS

© Oxford University Press 2011


Shape Number

• 1. Find the major axis and minor axis of the


boundary.
• 2. Using the major and minor axes, the sampling
boundary is constructed.
• 3. Resampling process is carried out and the
chain code is obtained.
• 4. Normalization is carried out by obtaining the
first difference.
• 5. The shape number is obtained by obtaining
the minimum magnitude of the first difference.

© Oxford University Press 2011


Fourier Descriptors

© Oxford University Press 2011


Run-length Code

© Oxford University Press 2011


Projections

© Oxford University Press 2011


Concavity Tree

© Oxford University Press 2011


COMPONENT LABELLING

© Oxford University Press 2011


Recursive Algorithm
• 1. Initialize the label L to -1.
• 2. Scan the image left to right and top to bottom.
• 3. Initialize a queue Q.
• 4. If the pixel value f(x, y) is greater than zero then insert (x, y) into the
queue.
• 5. If the queue is not empty, proceed to step 6; else proceed to step 12.
• 6. Remove the value at the front of the queue and assign it temporarily to
(s,t).
• 7. For all the neighbours of (s,t) do steps 8–11.
• 8. Recursively check the neighbour and assign the same label if they are
unlabelled.
• 9. Insert it into the queue.
• 10. Increment the label as L = L - 1. The labels are negative numbers for
distinguishing them from the pixel values.
• 11. Stop when all the pixels with the value ‘1’ are labelled.
• 12. Exit.

© Oxford University Press 2011


Sequential Algorithm

© Oxford University Press 2011


Sequential Algorithm

© Oxford University Press 2011


REGIONAL DESCRIPTORS

• Histogram (or Brightness) Features

© Oxford University Press 2011


© Oxford University Press 2011
Contd…

© Oxford University Press 2011


Additional Features

© Oxford University Press 2011


Contd…

© Oxford University Press 2011


Shape Features

• Shape is one of the most important


features of an object.

• Shapes can be generically classified for


further analysis into two types as
1.2D shapes 2. 3Dshapes

© Oxford University Press 2011


Geometrical Features

© Oxford University Press 2011


Area

© Oxford University Press 2011


Area

© Oxford University Press 2011


Perimeter

© Oxford University Press 2011


© Oxford University Press 2011
Boundary Box Area

© Oxford University Press 2011


Parameters

© Oxford University Press 2011


Spatial moments

© Oxford University Press 2011


Spatial moments

© Oxford University Press 2011


Contd…

© Oxford University Press 2011


Central Moments

© Oxford University Press 2011


Topological Features

© Oxford University Press 2011


Transform Features

© Oxford University Press 2011


Transform Features

© Oxford University Press 2011


Texture Features

© Oxford University Press 2011


Texture analysis

© Oxford University Press 2011


GLCM

© Oxford University Press 2011


© Oxford University Press 2011
Syntactic and Structural
Features

© Oxford University Press 2011


FEATURE SELECTION
TECHNIQUES

© Oxford University Press 2011


Principal component analysis
(PCA)

© Oxford University Press 2011


© Oxford University Press 2011
Summary

© Oxford University Press 2011


© Oxford University Press 2011

You might also like