Illumination Invariant Face Recognition By Expected Patch Log Likelihood

Abstract

Illumination is an important factor that impairs face recognition. Many algorithms have been proposed to solve the illumination problem. Most algorithms focus on one image information and only use local illumination change, to improve the effects of removing facial illumination. In this paper, we apply the Expected Patch Log Likelihood (EPLL) algorithm to extract illumination weight and we combine it with the Neighboring Radiance Ratio algorithm (NRR) to optimize the initial vector of the Gaussian mixture model, which makes full use of the redundant information in images. The experimental results on the extended Yale B and CMU PIE face databases show that the proposed algorithm can effectively eliminate the influence of illumination on face images and has a high robustness.