Face recognition using a non-zero-order correlation plane and a nonlinear joint transform correlator
Ayman Alfalou, Nesrine Ben-Haj-Yahia, M. Elbouz, M. S. Alam · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2012
For face recognition applications, it is necessary to have a robust discrimination system. In this paper, a new method for denoising the correlation plane and removing the zero-order term associated with alternate joint transform correlator (JTC) architectures such as the nonlinear JTC (NJTC) and nonlinear non-zero-order JTC (NNJTC) is proposed. The proposed technique is called nonlinear denoised JTC (NDJTC) which incorporates the attractive features of NJTC and NNJTC to ensure good compromise between discrimination and robustness while denoising the correlation plane. To investigate the performance of the proposed technique, various tests are performed using the PHPID face database and the results demonstrate excellent behavior of the proposed NDJTC. In this technique, the zero-order term and the impact of noise in the correlation plane were removed leading to increased robustness and discrimination ability of the NJTC. To confirm these results, different comparisons were performed using a NJTC and a NNJTC. Finally, a study using different levels of nonlinearity was conducted to find the best compromise between robustness and discrimination ability of the proposed method.