Threshold for Face classes in Face Recognition.

Date of Submission

December 2008

Date of Award

Winter 12-12-2009

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science

Department

Machine Intelligence Unit (MIU-Kolkata)

Supervisor

Murthy, C. A. (MIU-Kolkata; ISI)

Abstract (Summary of the Work)

Task of Face Recognition: Given still or video images of a scene, identify or verify one or more persons in the scene using a stored database of faces.The solution to the problem involves the segmentation of faces (face detection) from image of a scene, feature extraction from face region, recognition.A basic structure of a Face Recognition System is shown in the following diagram.Basic Structure of a generic Face Recognition System The main parts are typically face detection and face recognition which can it self be decomposed in normalization, feature extraction and classification steps.Face detection aims to determine whether or not there are any faces in the image and, if present, return the face location while the goal of face localization is to estimate the position of a single face.Feature extraction is to find a specific representation of the data that can highlight relevant information. At the feature extraction stage, the goal is to find an invariant representation of the face image. Usually, an image is represented by a high dimensional vector containing pixel values (holistic representation) or a set of vectors where each vector contains gray levels of a sub-image (local representation). There are different methods existing in the literature.i) Holistic Methodsii) Structural Matching Methodsiii) Hybrid Methodsi) Holistic methods: These methods use the whole face region as the raw input to the system for extracting the features of a face image.Principal Component Analysis:One of the feature extraction techniques, based on Principal Component Analysis (PCA), was first used for face recognition by Turk and Pentland [4]. The aim of PCA is to find a representation of the data minimizing the reconstruction error. The PCA finds the orthogonal directions that account for the highest amount of variance. The data is then projected into the subspace spanned by these directions. In practice, the principal component axes are the eigenvectors of the covariance matrix of the data. The corresponding eigen values indicate the proportion of variance of the data projections along each direction.Linear Discriminant Analysis:Another feature extraction method used in face recognition is based on Linear Discriminant Analysis (LDA), also known as Fisher Discriminant Analysis [5].The LDA subspace holds more discriminant features than the PCA subspace. LDA finds a subspace in which the variability is maximized between different class data, and at the same time where variability in the same class data (face images of the same identity) is minimum.We define the within-class scatter matrix as Sw and between-class scatter matrix as Sb .The goal is to maximize the between-class measure while minimizing the withinclass measure. One way to do this is to maximize the ratio det(Sb)/det(Sw.)Intuitively, for face recognition, LDA should outperform PCA because it inherently deals with class discrimination. However, Martinez and Kak [5] have shown that PCA might outperform LDA when the number of samples per class is small.ii) Structural matching methods: In these methods, local features such as the eyes, nose, and mouth are first extracted and their locations and local statistics (geometric and/or appearance) are fed into a structural classifier.

Comments

ProQuest Collection ID: http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqm&rft_dat=xri:pqdiss:28843229

Control Number

ISI-DISS-2008-225

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

DOI

http://dspace.isical.ac.in:8080/jspui/handle/10263/6383

This document is currently not available here.

Share

COinS