Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 
Finding Near Optimum Colour
Classifiers : Genetic Algorithm-Assisted
Fuzzy Colour Contrast Fusion using
Variable Colour Depth
A thesis presented to the
Institute of Information and Mathematical Sciences
in partial fulfillment of the requirements for the degree of
Master of Science in Computer Science
at
Massey University, Albany, Auckland, New Zealand
By
Heesang Shin
May 2009

c© Copyright 2009
by
Heesang Shin
iii
iv
Abstract
This thesis presents a complete self-calibrating illumination intensity-invariantcolour classification system. We extend a novel fuzzy colour processing tech-
nique called Fuzzy Colour Contrast Fusion (FCCF) by combining it with a Heuristic-
assisted Genetic Algorithm (HAGA) for automatic fine-tuning of colour descriptors.
Furthermore, we have improved FCCF’s efficiency by processing colour channels at
varying colour depths in search for the optimal ones. In line with this, we intro-
duce a reduced colour depth representation of a colour image while maintaining
efficient colour sensitivity that suffices for accurate real-time colour-based object
recognition. We call the algorithm Variable Colour Depth (VCD) and we propose
a technique for building and searching a VCD look-up table (LUT). The first part
of this work investigates the effects of applying fuzzy colour contrast rules to vary-
ing colour depths as we extract the optimal rule combination for any given target
colour exposed under changing illumination intensities. The second part introduces
the HAGA-based parameter-optimisation for automatically constructing accurate
colour classifiers. Our results show that for all cases, the VCD algorithm, combined
with HAGA for parameter optimisation improve colour classification via a pie-slice
colour classifier.For 6 different target colours, the hybrid algorithm was able to
yield 17.63% higher overall accuracy as compared to the pure fuzzy approach. Fur-
thermore, it was able to reduce LUT storage space by 78.06% as compared to the
full-colour depth LUT.
v
vi
Preface
Some merits of this work has already been recognised, published and submit-ted.
Lecture Notes in Computer Science, Springer-Verlag (accepted, to appear in 2009) :
Variable Colour Depth Look-Up Table Based on Fuzzy Colour Processing, Heesang
Shin and Napoleon H. Reyes, In Proceedings of ICONIP 2008
Memetic Computing Journal, Springer (submitted in 2009) : Finding Near Opti-
mum Colour Classifiers : Genetic Algorithm-Assisted Fuzzy Colour Contrast Fusion
using Variable Colour Depth, Heesang Shin and Napoleon H. Reyes
vii
viii
Acknowledgements
First , I would like to thank Dr. Napoleon Hamoay Reyes, without his insightand guidance I wouldn’t finish this thesis. That’s why I used pronoun we
instead of I, he deserves it. I also equally thanks to my wife Kristen KyungEun,
it wouldn’t appropriate to list just number of things to express her total support.
Finally I dedicate this thesis to my elder son Daniel JinKyu, without him I wouldn’t
able to return to study after long years in industry.
ix
x
Contents
Abstract v
Preface vii
Acknowledgements ix
1 Research Description 1
1.1 Overview of the Current State of Technology . . . . . . . . . . . . . 1
1.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 General Objective . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Specific Objectives . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Scope and Limitations of Research . . . . . . . . . . . . . . . . . . 3
1.4 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Theoretical Framework 5
2.1 Colour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Colour Space Models . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 CIE XYZ Colour Space and CIE 1931 Chromaticity Diagram 7
2.2.2 RGB Colour Model . . . . . . . . . . . . . . . . . . . . . . . 10
xi
2.2.3 CMY and CMYK Colour Models . . . . . . . . . . . . . . . 11
2.2.4 HSI Colour Model . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Colour Image Formation . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Colour Separation Mechanism . . . . . . . . . . . . . . . . . 16
2.4 Colour Representation in Binary . . . . . . . . . . . . . . . . . . . . 18
2.4.1 Colour Depth . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 Review of Related Literature 21
3.1 Colour Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.1 Indexing Via Color Histograms . . . . . . . . . . . . . . . . 21
3.1.2 A Robust and Fast Color-Extracting using a Look up Table 23
3.1.3 Color recognition . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.4 Real-time, adaptive color-based robot vision . . . . . . . . . 24
3.1.5 A Fast Algorithm for Color Image Segmentation . . . . . . . 25
3.1.6 Towards a calibration-free robot: The ACT algorithm for au-
tomatic online color training . . . . . . . . . . . . . . . . . . 26
3.1.7 Automatic On-Line Color Calibration using Class-Relative
Color Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.8 Adaptive recognition of color-coded objects in indoor and out-
door environments . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.9 Mean-shift-based color tracking in illuminance change . . . . 30
3.1.10 Robust color classification using fuzzy rule-based Particle Swarm
Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
xii
3.2.1 Knowledge-Based Fuzzy Color Processing . . . . . . . . . . . 33
3.3 Fuzzy Colour Contrast Fusion (FCCF) . . . . . . . . . . . . . . . . 35
3.3.1 Dynamic Colour Object Recognition Using Fuzzy Logic . . . 35
3.3.2 Identifying Colour Objects with Fuzzy Colour Contrast Fusion 40
3.3.3 Hybrid Fuzzy Colour Processing and Learning . . . . . . . . 43
3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Central Thesis 49
4.1 Variable Colour Depth . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1.1 Look-up Table (LUT) . . . . . . . . . . . . . . . . . . . . . 58
4.1.2 LUT Building Algorithm . . . . . . . . . . . . . . . . . . . . 61
4.1.3 LUT Query Algorithm . . . . . . . . . . . . . . . . . . . . . 61
4.1.4 General Variable Colour Depth - FCCF System Architecture 62
4.2 Fuzzy-Genetic Colour Classifier Search . . . . . . . . . . . . . . . . 63
4.2.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.2 General Architecture . . . . . . . . . . . . . . . . . . . . . . 66
4.2.3 Chromosome Design . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4 Fitness Function . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5 Experiments and Analysis 69
5.1 Test Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.1.1 Assessment Method . . . . . . . . . . . . . . . . . . . . . . . 69
5.1.2 Reference Result . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2 Variable Colour Depth with CCRE . . . . . . . . . . . . . . . . . . 70
xiii
5.2.1 Search Strategy . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2.2 Colour Classification Results of Full 24-bit Colour Depth vs.
Variable Colour Depth . . . . . . . . . . . . . . . . . . . . . 71
5.2.3 Colour Contrast Rules and Scores . . . . . . . . . . . . . . . 72
5.2.4 Colour Contrast Rule Clustering . . . . . . . . . . . . . . . . 73
5.2.5 Colour Pixel Clustering . . . . . . . . . . . . . . . . . . . . . 74
5.2.6 Reductions in Memory Usage . . . . . . . . . . . . . . . . . 83
5.2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3 Fuzzy-Genetic Colour Calibration . . . . . . . . . . . . . . . . . . . 83
5.3.1 Fuzzy-Genetic Colour Calibration Parameters and Scores . . 86
5.3.2 Colour Contrast Rule Component Distribution . . . . . . . . 86
5.3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6 Conclusions 99
6.1 Suggestions for Future Work . . . . . . . . . . . . . . . . . . . . . . 100
Bibliography 101
Appendices 105
A Proposed System : FCCF Suite 105
A.1 Licences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
A.2 Software Integration . . . . . . . . . . . . . . . . . . . . . . . . . . 106
A.2.1 Qt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
A.2.2 OpenCV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
A.2.3 QextSerialPort . . . . . . . . . . . . . . . . . . . . . . . . . 106
xiv
A.2.4 TinyXML . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
A.2.5 GAlib . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
A.3 Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
A.3.1 FCCF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
A.3.2 Cross-Platform Compatibility . . . . . . . . . . . . . . . . . 108
A.3.3 Video Capture . . . . . . . . . . . . . . . . . . . . . . . . . 109
A.3.4 Real-Time Object Tracking . . . . . . . . . . . . . . . . . . 109
A.3.5 GUI System . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
A.3.6 Robot Control . . . . . . . . . . . . . . . . . . . . . . . . . . 109
A.4 Test-Bed Hardware Specifications . . . . . . . . . . . . . . . . . . . 110
xv
List of Tables
1 Quality Criteria For Good And Poor Welding Spots. From Knowledge-
Based Fuzzy Color Processing, 2004 . . . . . . . . . . . . . . . . . . 34
2 Colour Descriptors for the Target Colours [1]. . . . . . . . . . . . . 42
3 Colour Contrast Rules for Each rg-chromaticity, YUV and, HSI Colour
Spaces [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4 False positive and true positive rates for the Colour Contrast Fusion
Algorithm in rg-Chromaticity, YUV, and HSI Colour Spaces [1]. . . 42
5 Sample Variable Colour Depth Representations of the Normalised
Colour Component Values 0.8 Red, 0.5 Green, and 1.0 Blue . . . . 50
6 Comparisons of Colour Classification Result between Indexed and
VCD LUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7 Colour Classification Definition . . . . . . . . . . . . . . . . . . . . 70
8 Colour Classification Results of Full 24-bit Colour Depth vs. Variable
Colour Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
9 Colour Contrast Rule Distribution . . . . . . . . . . . . . . . . . . . 73
10 Fuzzy-Genetic Colour Calibration Experiment Configuration . . . . 91
11 Fuzzy-Genetic Colour Calibration Result for Yellow . . . . . . . . . 92
12 Fuzzy-Genetic Colour Calibration Result for Green . . . . . . . . . 93
13 Fuzzy-Genetic Colour Calibration Result for Pink . . . . . . . . . . 94
xvi
14 Fuzzy-Genetic Colour Calibration Result for Purple . . . . . . . . . 95
15 Fuzzy-Genetic Colour Calibration Result for Violet . . . . . . . . . 96
16 Fuzzy-Genetic Colour Calibration Result for Light Blue . . . . . . . 97
17 Colour Contrast Rule Component Distribution . . . . . . . . . . . . 98
18 Colour Pixel Distribution Changes after FCCF Applied in 6 Target
Colours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
19 Colour Classification and Contrast Angle Difference Between Fuzzy-
Genetic Optimised Solution and Manual Calibrated Solution (size
difference represents the angle difference relative to the base) . . . . 98
xvii
List of Figures
1 Electromagnetic spectrum with Light Highlighted Picture created by
Philip Ronan from Wikipedia . . . . . . . . . . . . . . . . . . . . . 6
2 Leaf Reflects Green Wavelength on The Surface to be Perceived . . 7
3 Schematic Diagram of the Human Eye and Cross Section View of
Retina. Schematic created by Rhcastilhos from Wikipedia . . . . . . 8
4 An Optical Illusion. Square A is Exactly the Same Shade of Grey as
Square B. Picture created by Adrian Pingstone, based on the original
created by Edward H. Adelson [2] . . . . . . . . . . . . . . . . . . . 9
5 Simplified Human Cone Response Curve and Corresponding CIE
XYZ Colour Matching Function . . . . . . . . . . . . . . . . . . . . 10
6 The CIE 1931 Colour Space Chromaticity Diagram. Picture created
by Sakurambo from Wikipedia, based on the original created by CIE [3] 11
7 Schematic of the RGB Colour Cube . . . . . . . . . . . . . . . . . . 12
8 Flatten Schematic of the RGB Colour Cube . . . . . . . . . . . . . 13
9 The HSI-Colour Space . . . . . . . . . . . . . . . . . . . . . . . . . 14
10 A Transistor-Level Schematic of A Three-Pixel, Photodiode-Based
Active Pixel Sensor. Diagram created by Gargan from Wikipedia . . 16
xviii
11 A Philips Type Trichroic Beam Splitter Prism Schematic, With a
Different Colour Separation Order Than the Assembly Shown in the
Photo. The Red Beam Undergoes Total Internal Reflection at the
Air Gap, While the Other Reflections are Dichroic. Diagram created
by Gargan from Wikipedia . . . . . . . . . . . . . . . . . . . . . . . 17
12 The Bayer Arrangement of Colour Filters on Image Sensor Array.
Diagram created by Cburnett from Wikipedia . . . . . . . . . . . . . 18
13 Profile/Cross-Section of Bayer Filter Layered Sensor. Diagram cre-
ated by Cburnett from Wikipedia . . . . . . . . . . . . . . . . . . . . 18
14 Two-Layer Pyramid Structure for Each Colour Component [4]. . . . 25
15 Block Diagram of the Colour Classification System [5]. . . . . . . . 28
16 Captured Panoramic Images with PID Controller under Four Light
Conditions [6]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
17 Plots of YUV Colour Distribution Indoors [6]. . . . . . . . . . . . . 31
18 Plots of YUV Colour Distribution in Outdoor Environment [6]. . . 31
19 Adjusting RGB Colour Value by F-number [7]. . . . . . . . . . . . . 32
20 The HSL color Space Mapped to a Sphere, with Corner Cut-Away
Shown. Figure created by SharkD from Wikipedia . . . . . . . . . . 33
21 The Demension H [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . 33
22 Dimensions L and S [8]. . . . . . . . . . . . . . . . . . . . . . . . . 34
23 Regions of Interest of a Resistance Spot Welding Joint [9]. . . . . . 35
24 Fuzzy Set, Defined Over the HS-Colour Space [9]. . . . . . . . . . . 35
25 Colour Contrast and Classification System Architecture [10]. . . . . 36
26 rg-Chromaticity Colour Space with Origin Shift Position [10]. . . . 37
27 Pie-Slice Colour Decision Region in Modified rg-Chromaticity Colour
Space [10]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
xix
28 Colour Contrast Enhance Operator (Sigmoid / Logistic Function) [10]. 39
29 Colour Contrast Degrade Operator (Logit Function) [10]. . . . . . . 40
30 Test Image with Pink Colour Patches in the Middle [10]. . . . . . . 40
31 Colour Classified Image. True Positive Pixels are in Light Blue, False
Positive Pixels are in Blue Colours. . . . . . . . . . . . . . . . . . . 41
32 Results of Applying Colour Contrast Fusion in rg-Chromaticity, YUV,
and HSI Colour Spaces [1]. . . . . . . . . . . . . . . . . . . . . . . . 43
33 The MPCL Algorithm [11]. . . . . . . . . . . . . . . . . . . . . . . 44
34 Extracted Object Colour Pixels From Two Consecutive Frames and
Corresponding Colour Classification Variable Changes [11]. . . . . . 44
35 Normalised Input Values from Various Colour Depth . . . . . . . . 51
36 Enlarged Section of Normalised Input Values from Various Colour
Depth Input Values from 1/8 to 2/8 . . . . . . . . . . . . . . . . . . 52
37 Examples of Colour Depth Reduction; (a) Original 24-bit RGB Im-
age; (b) Reduced 15-bit RGB Image from (a); (c) 8-bit Gray-Scale
Image from (a); (d) 4-bit Gray-Scale Image from (c). . . . . . . . . 53
38 Examples of Colour Depth Reduction; (a) Original 24-bit RGB Im-
age, 55,880 Colours were Used; (b) Areas Where Colour Differences
are Visible; (c) 16-bit RGB Image, 4,391 Colours were Used; (d)
15-bit RGB Image 2,814 Colours were Used. . . . . . . . . . . . . . 54
39 Comparisons of Colour Contrast Enhancement (1X Mode) Results
Using Various Colour Depth Representations. . . . . . . . . . . . . 55
40 Comparisons of Colour Contrast Enhancement (2X Mode) Results
Using Various Colour Depth Representations. . . . . . . . . . . . . 55
41 Comparisons of Colour Contrast Enhancement (3X Mode) Results
Using Various Colour Depth Representations. . . . . . . . . . . . . 55
xx
42 Comparisons of Colour Contrast Degradation (1X Mode) Results Us-
ing Various Colour Depth Representations. . . . . . . . . . . . . . . 56
43 Comparisons of Colour Contrast Degradation (2X Mode) Results Us-
ing Various Colour Depth Representations. . . . . . . . . . . . . . . 56
44 Comparisons of Colour Contrast Degradation (3X Mode) Results Us-
ing Various Colour Depth Representations. . . . . . . . . . . . . . . 56
45 Examples of Normalised Outputs Produced by the Lower Colour
Component Depths. . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
46 RGB Colour Space in Variable Colour Depth of 3-8-8. . . . . . . . . 58
47 Comparisons between Standard Indexed LUT and VCD LUT. . . . 59
48 Shades of Colour between Blue and Pink. . . . . . . . . . . . . . . . 61
49 Variable Colour Depth Look-Up Table Construction Architecture . 63
50 Variable Colour Depth Look-Up Table for Real-Time Processing . . 64
51 The Shaded Pie-Slice, Covered by Arc L Represents The Area of
Interest. Due to the Discretisation of the Angles, the Area of Interest
Could Only be Approximated by a Smaller Pie-Slice, Covered by
Angle θ. The Dotted Lines Indicate the Discretisation of Angles. . . 65
52 Fuzzy-Genetic Colour Classifier Search Architecture . . . . . . . . . 66
53 Chromosome Design . . . . . . . . . . . . . . . . . . . . . . . . . . 67
54 Mapping of all the Best Colour Contrast Rule Combinations for all
Colour Depth Values and for each Target Colour . . . . . . . . . . . 74
55 Mapping of the Best Colour Contrast Rule Combinations for the
Optimal Colour Depths for each Target Colour. Positive Number In-
dicates Contrast Enhancement and Level of Contrast Application; 0
for No Operation, while a Negative Number Denotes Contrast Degra-
dation. * n Indicates Number of Occurrences. . . . . . . . . . . . . 75
xxi
56 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Light
Blue Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
57 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Light
Blue Objects with FCCF . . . . . . . . . . . . . . . . . . . . . . . . 76
58 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Yellow
Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
59 Enlarged Colour Pixel Clustering on rg-Hue / rg-Saturation Chart
for Yellow Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
60 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Green
Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
61 Enlarged Colour Pixel Clustering on rg-Hue / rg-Saturation Chart
for Green Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
62 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Pink
Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
63 Enlarged Colour Pixel Clustering on rg-Hue / rg-Saturation Chart
for Pink Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
64 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Purple
Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
65 Enlarged Colour Pixel Clustering on rg-Hue / rg-Saturation Chart
for Purple Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
66 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Violet
Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
67 Enlarged Colour Pixel Clustering on rg-Hue / rg-Saturation Chart
for Violet Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
68 Colour Pixel Clustering on rg-Hue / rg-Saturation Chart for Light
Blue Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
xxii
69 Enlarged Colour Pixel Clustering on rg-Hue / rg-Saturation Chart
for Light Blue Objects . . . . . . . . . . . . . . . . . . . . . . . . . 82
70 Contrast Rule Component Distribution for Yellow . . . . . . . . . . 87
71 Contrast Rule Component Distribution for Green . . . . . . . . . . 87
72 Contrast Rule Component Distribution for Pink . . . . . . . . . . . 88
73 Contrast Rule Component Distribution for Purple . . . . . . . . . . 88
74 Contrast Rule Component Distribution for Violet . . . . . . . . . . 89
75 Contrast Rule Component Distribution for Light Blue . . . . . . . . 89
76 A Screen-Shot of the Proposed System. . . . . . . . . . . . . . . . . 110
xxiii
List of Algorithms
1 The Adaptive Colour-Based Robot Vision Algorithm [12]. . . . . . . 24
2 Look-Up Table Building Algorithm . . . . . . . . . . . . . . . . . . . 41
3 CCRE(image, targetbounds), Scoring Formula [11]. . . . . . . . . . . 45
4 Variable Colour Depth LUT Build Algorithm . . . . . . . . . . . . . 62
5 Variable Colour Depth LUT Query Algorithm . . . . . . . . . . . . . 63
xxiv