Title Page
Abstract
Contents
I. Introduction 9
II. Background 13
2.1. Color Balancing 13
2.2. Illumination Compensation 16
III. Illumination Compensated Face Detection in the Compressed Domain 19
3.1. DCT Parameters Extraction 21
3.2. Color Balancing and Illumination Compensation 21
3.3. Human Skin Color Model Matching 23
3.4. Face Candidate Extraction 24
3.5. Face Candidate Verification 26
IV. Experiments 29
V. Conclusions 36
References 38
Table 1. Numerical data for the experiment. 34
Figure 1. CIE Chromaticity Diagram [18] 15
Figure 2. Effects of the illumination compensation 18
Figure 3. represents the overview of the proposed illumination compensated 19
Figure 3. The overview framework of the proposed face detection system. 20
Figure 4. Results of Color Balancing. 22
Figure 5. Results of Illumination Compensation. 23
Figure 6. Skin color region matching 24
Figure 7. The face template 25
Figure 8. The grouping of the DCT coefficients. 27
Figure 9. Results of the color balancing, illumination compensation and proposed method applied with CMU PIE database. 31
Figure 10. TP and FP rates of the four experiments 33
Figure 11. The output of each stage in the proposed system. 35