Unconstrained Face Recognition

Although face recognition has been actively studied over the past decade, the state-of-the-art recognition systems yield satisfactory performance only under controlled scenarios and recognition accuracy degrades significantly when confronted with unconstrained situations due to variations such as illumintion…


1.1 Overview
1.1.1 Biometric perspective
1.1.2 Experimental perspective
1.1.3 Theoretic perspective
1.2 Unconstrained Face Recognition
1.2.1 Face recognition under variations
1.2.2 Face recognition via kernel learning
1.2.3 Face tracking and recognition from videos
2 Generalized Photometric Stereo
2.1 Principle of Generalized Photometric Stereo
2.1.1 Literature review and proposed approach
2.1.2 Setting and constraints
2.1.3 Separating illumination
2.1.4 Recovering class-specific albedos and surface normals
2.2 Face Recognition across Illumination
2.2.1 Literature review and proposed approach
2.2.2 Bootstrap set
2.2.3 Recognition experiments
2.3 Appendix
3 Illuminating Light Field
3.1 Principle of Illuminating Light Field
3.1.1 Literature review
3.1.2 Pose-invariant identity signature
3.1.3 Illumination- and pose-invariant identity signature
3.1.4 Learning algorithms
3.2 Face Recognition across Illumination and Poses
3.2.1 PIE database and recognition setting
3.2.2 Recognition performance
3.2.3 Comparisons
4 Probabilistic Kernel Principal Component Analysis
4.1 Reproducing Kernel Hilbert Space (RKHS)
4.2 Probabilistic Analysis of Kernel Principal Components
4.2.1 Kernel principal component analysis
4.2.2 Theory of PKPCA
4.3 Mixture Modeling of Probabilistic Kernel Principal Components
4.3.1 Theory of mixture of PKPCA
4.3.2 Why mixture of PKPCA?
4.4 Classification
4.4.1 PKPCA or mixture of PKPCA classifier
4.4.2 Experiments
4.5 Appendix
5 Probability Distances in Reproducing Kernel Hilbert Space
5.1 Probabilistic Distances in Rd
5.2 Mean and Covariance Marix in RKHS
5.2.1 First- and second-order statistics
5.2.2 Covariance matrix approximation
5.3 The Probabilistic Distances in RKHS
5.3.1 The Chernoff distance and the Bhattarchayya distance
5.3.2 The KL divergence and the symmetric divergence
5.3.3 The Patrick-Fisher distance
5.3.4 Limiting behavior
5.3.5 Kernel for set
5.4 Experimental Results
5.4.1 Synthetic examples
5.4.2 Face recognition from a group of images
6 Adaptive Visual Tracking
6.1 Related Literature
6.1.1 Visual tracking
6.1.2 Particle filter
6.2 Appearance-Adaptive Models
6.2.1 Adaptive observation model
6.2.2 Adaptive state transition model
6.2.3 Handling occlusion….

Author: Zhou, Shaohua

Source: University of Maryland

Keywords: Project Reports, Dissertation, Thesis

Leave a Comment