Face Recognition using FSS-DSOP for Small Sample Size Problem with Illumination Variations

In this paper we propose a method termed as Face-Specific Subspace DCT Sign Only Product (FSS–DSOP), for face recognition. The proposed method is based on Face-Specific Subspace (FSS) technique using Discrete Cosine Transform (DCT) sign only product as similarity measure. The proposed approach aims to improve the recognition rate of the FSS technique under small size scenario with illumination variations by making use of DCT sign-only inner product similarity measure for classification. Results obtained from a set of experiments using small training sets sampled from illumination subset of extended Yale-B and PIE face database indicate that the proposed methodology achieves better recognition rate when compared to FSS.