Face Recognition with Modular Two Dimensional PCA under Uncontrolled Illumination Variations

Abstract

Person authenticaton using faces became one of the most popular security approaches for the last three decades. From the literature it is found that perofrmance of most of the methods used in recognition was limited due to uncontrolled conditions like illumination and pose variations. In this work, to address the limitations of uncontrolled environment, Modular two-dimensional Principle Component Analysis (M2D-PCA) is proposed. In this approach, the input image is partitioned into four equal segments and then Histogram Equalization is applied to reduce illumination impact caused due to varying lightening conditions. Then M2D-PCA algorithm is applied parallel on each segment and then all features extracted from the segments are fused with wieghted summation. Experiments are carried out on bench mark datasets like extended Yale database B, ORL and AR database. Results of the proposed approach produced good recognition rate with low computational time against various illumination environments.