Post-processed LDA for face and palmprint recognition: What is the rationale

Abstract

Linear discriminant analysis (LDA)-based methods have been very successful in face and palmprint recognition. Recently, a class of post-processing approaches has been proposed to improve the recognition performance of LDA in face recognition. In-depth analysis, however, has not been presented to reveal the effectiveness of the post-processing approach. In this paper, we first investigate the rationale of the post-processing approach using a Gaussian function, and demonstrate the mutual relationship between the post-processing approach and the image Euclidean distance (IMED) method. We further extend the post-processing approach to palmprint recognition and use the FERET face and the PolyU palmprint databases to evaluate the post-processed LDA method. Experimental results indicate that the post-processing approach is effective in improving the recognition rate for LDA-based face and palmprint recognition.