1. Discovery of Human-Competitive Image Texture Feature Extraction Programs Using Genetic Programming
By Brian Lam and Vic Ciesielski
RMIT University
School of Computer Science and Information Technology

2. What is texture ?
Texture can be considered to be repeating patterns of local
variation of pixel intensities.
Brodatz Textures
Vistex Textures

3. Human Invented Algorithms
Texture feature extraction algorithms can be grouped as follows*
Statistical
Geometrical
Model based
Signal Processing
*Tuceryan and Jain, “Texture Analysis” in The Handbook of Pattern Recognition and Computer Vision,
World Scientific, 2nd edn., 1998

4. Statistical Methods Local features Autoregressive
Galloway – run length matrix
Haralick – co-occurrence matrix
Unser
Sun and Wee
Amadasun
Dapeng
Amalung

5. Local Features Grey level of central pixels
Average of grey levels in window
Median
Standard deviation of grey levels
Difference of maximum and minimum grey levels
Difference between average grey level in small and large windows
Sobel feature
Kirsch feature
Derivative in x window
Derivative in y window
Diagonal derivatives
Combine features

6. Haralick Features
First transform pixels into a co-occurrence matrix then calculate a (large) number of statistical features from the matrix.

9. Geometric Methods Chen’s geometric features
First threshold images into binary images of n grey levels
Then calculate statistical features of connected areas.

10. Model Based Methods
These involve building mathematical models to describe textures.
Markov random fields
Fractals 1
Fractals 2

11. Signal Processing Methods
These methods involve transforming original images using filters and calculating the energy of the transformed images.
Law’s masks
Laines – Daubechies wavelets
Fourier transform
Gabor filters

12. Research Questions
How do we use GP to evolve texture feature extraction programs ?
- Inputs
- Functions
- Fitness evaluation
Can GP generate human competitive feature extraction programs ?

13. Texture Classification
Classical Approach
Our Approach
Feature Extraction invented
by human
Feature Extraction discovered
by GP
Extract Features from Vistex
Extract Features from Vistex
Training data
Testing Data
Training data
Testing Data
Classifier
Classifier
Test on testing data
Test on testing data

14. Discovering Programs Using GP
Evolve feature extraction programs
Learning Data (Brodatz)
Extract Features
Evaluate Fitness
Feature Extraction programs discovered
by GP

15. Data Set Definitions
Learning set: 13 Brodatz textures used to evolve 78 programs (80 of 64 x 64 images in each).
Training set: 15 Vistex textures used to train classifier (32 of 64 x 64 images in each ).
Testing set: 15 Vistex textures used to test classifier (64 of 64 x 64 images in each).
*Wagner T, “Texture Analysis” in Handbook of Computer Vision and Applications, Academic Press, 1999

16. GP Configuration Brodatz Texture Images 256 inputs Histogram Values
Image size
64 x 64
GP System
Operator : plus
Fitness Evaluation : Overlap between clusters
Texture Feature Extraction Programs

17. Feature Space for Two Textures

18. Histograms of Class 1 and Class 2 Learning Set Textures
Evolved program :
X *X *X117 +
X *X132 + X133 +
2*X143 + X151 +X206 +
X238 +3*X242 + X254

19. Results Accuracy % GP features
*Wagner T, “Texture Analysis” in Handbook of Computer Vision and Applications, Academic Press, 1999

20. RESULTS 2 Industrial inspection problem Classification of Malt Images
Our GP features slightly more accurate than Haralick features

21. Conclusions
GP can generate feature extraction algorithms that are competitive with human developed algorithms.
Evolved programs are fast compared with some of the human derived ones.

22. Inputs Histograms Pixels 16 x 16 = 256 inputs64 16 16 64
16 x 16 = 256 inputs
256 grey levels = 256 inputs

23. Other GP Parameters Generation : 200 Mutation rate : 0.28
Cross-over rate : 0.78
Elitism : 0.02