Event Details
An enhanced and extended principal component analysis (PCA) for face recognition
Presenter: Ana-Maria Sevcenco
Supervisor: Dr. Wu-Sheng Lu
Date: Mon, August 23, 2010
Time: 14:00:00 - 15:00:00
Place: EOW 230
ABSTRACT
ABSTRACT
Numerous PCA-based methods have been developed to improve PCA performance since the original work of Turk and Pentland (1991). A two-dimensional (2-D) approach for face recognition purpose was proposed by Yang et al. in 2004. It consists of an image projection technique in which images are treated as matrices instead of vectors as in the original PCA. This method leads not only to better recognition rates, but also to improved computationally efficiency.
We take a closer look at the 2-D PCA algorithm and propose an extended 2-D PCA (E-2DPCA) algorithm that utilizes a pair of covariance matrices to extract both row-related and column-related features of facial images. In conjunction with our extension is a new criterion for face classification. We give a short overview of several face recognition algorithms (PCA, 2-D PCA and sparse representation) and compare their performance to that of the proposed face recognition algorithm. To further enhance the performance of E-2DPCA algorithm, a pre-processing technique referred as PHM is applied to obtain the PHM E-2DPCA face recognition system.