Contents

• List of Figures
• List of Tables
• Introduction
  • Problem statement
  • Structure of thesis
  
  
• State of the art
  • Appearance Based Methods
    • Approach of T. Kato, Y. Ninonmiya and I. Masaki [17]
    • Approach of V.S. Petrovic, T.F. Cootes [19]
  • Model Based Methods
  • Summary
  
  
• Proposed method
  • Structure of Proposed Method
    • Frontal Mask Localization and Geometric Image Normalization
    • Feature Extraction and Photometric Normalization
    • Classification
  • Detailed description of components
    • Distinctive Image Features from Scale-Invariant Keypoints -- SIFT
    • Error function
    • Classifiers used
  
  
• Experiments
  • Database description
    • Car images
    • Truck images
  • Parameters
    • SIFT representation
    • Used classifiers
  • k-NN performed on car database
    • Optimal SIFT topology
    • SIFT representation
    • Euclidean versus EMD error function
    • Best k for k-NN
  • k-NN Performed on truck database
    • SIFT topology
    • SIFT representation comparison
    • EMD vs. Euclidean distance
    • Best k for k-NN
  • Fisher's linear discriminator (FLD) performed on car images
    • Optimal dimension
  • Fisher's linear discriminator (FLD) performed on truck images
    • Optimal dimension
  
  
• Experimental validation
  • Sensitivity to learning set reduction
  • Sensitivity to image noise
  • Sensitivity to image blur
  • k-NN stability on truck images
    • Sensitivity to learning set reduction
    • Sensitivity to image blur
    • Sensitivity to image added noise
  • k-NN stability on car images
    • Sensitivity to learning set reduction
    • Sensitivity to image blur
    • Sensitivity to image added noise
  
  
• Conclusion
  • Discussion and further research
  
  
• Software design
  • Architecture overview
  • Program flow
  • Detailed modules design
    • Feature extraction module
    • Error module
    • Classification module
  
  
• Index
• Bibliography