Partitioning-based face recognition using PCA, LDA and ICA

Abstract

Partitioning approaches for facial images are presented for the improvement of the recognition performance of appearance-based statistical methods Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Independent Component Analysis (ICA). A face image is partitioned into several equal-width vertical or horizontal segments and a multiple classifier based divide-and-conquer approach is used to combine the recognition results associated with these segments to recognize the whole face. The experiments demonstrate that vertical and horizontal partitioning result in a better recognition performance compared to the performance results of the holistic PCA, LDA and ICA methods.