Contents

• Contents
• List of Figures
• List of Tables
• Introduction
  • Fundamental Issues in Face Recognition
  • Current Vision Systems for Face Recognition
  • The Proposed Approach
  • Structure of the Thesis
• Perceptual Contrast, Symmetry and Scale
  • Biological and Psychological Motivation
  • Low Level Filtering for Perceptually Interesting Objects
  • Edge Detection
    • The Sobel Operator
    • Deriche Edge Detection
    • Edge Data Enhancement
  • Multi-Scale Analysis
  • Real-Time Symmetry Transform
    • Lines of Symmetry
    • Intersection of Lines of Symmetry
    • Real-Time Implementation
    • Dark and Bright Symmetry
    • Application
  • Selective Symmetry Detection for Precise Blob Identification
    • Semi-Elliptical Sampling Regions
    • Projecting Edges onto the Template
    • Symmetric Enclosure
    • Application: Symmetric Enclosure versus Template Matching
• Face Detection and Localization
  • Face Localization
    • Face Blob Localization
    • Face Contour Estimation
  • Eye Localization
    • Detecting Eye Regions
    • Geometrical Tests
    • Rotation Transformation for Mouth and Nose Detection
  • Mouth Localization
    • Horizontal Symmetry Projection
    • Limb Extraction
    • Limb Length and Intensity Variance
  • Nose Localization
    • Vertical Signatures
  • Iris Localization
  • Improving and Filtering Localization
• Face Normalization and Recognition
  • 3D Face Data for Normalization
  • Generating the Average 3D Face
  • Inverse 3D Projection
    • P3P - Fitting the 3D Model to a 2D Image
    • Selecting the Optimal P3P for Deformation
    • Texture Mapping
    • Occlusion and Back Surfaces
    • Mirroring Along the Mid-Line
  • Synthesizing 2D Images with a Texture Mapped 3D Model
    • Synthesizing Segmented Mug-Shot Images for Recognition
  • Shading and Lighting Normalization
    • Histogram Fitting
    • Selecting a Target Histogram
    • Windowing Histogram Analysis
    • Beards and Hair
    • Gradated Histogram Fitting
  • Typical Normalization Results
  • Karhunen-Loeve Decomposition for Statistical Recognition and Detection
    • Computing Eigenvalues and Eigenvectors
    • Encoding Face Images with a Linear Combination Key
    • Decoding a Key into an Image
    • Varying n for Speed or Resolution
    • Using KL as a Faceness Detector
    • Nose Localization Revisited
    • Discarding Non-Faces before the Recognition Stage
    • Face Recognition with Distance Measures
  • Synopsis
• Implementation and Testing
  • Implementation
    • System Overview
    • Performance and Code Efficiency
    • Graphical User Interface
  • Testing
    • Localization Test
    • Recognition Test
    • Sensitivity Analysis
• Conclusions
  • Contributions
  • Direction of Future Work
• Bibliography