Illumination Invariant Face Recognition with Particle Swarm Optimization

In face recognition, the illumination variation problem in uncontrolled environments has gained some research activities. Although the quotient image based methods are reported to be a simple yet practical technique in face recognition, these methods could not satisfactorily maximize the ratios of between-class and within-class scatter and may not effectively be used for the illumination variation problem directly. In this paper, we proposed a new approach, termed as PSO-SQI, for the illumination variation problem. For illumination normalization under varying lighting conditions, our method uses the PSO-based feature selection in the Quotient face images to maximize the ratios of between-class and within-class scatter. Compared with the traditional SQI based approach in Yale Face database B, the experimental results show that our algorithms can significantly improve the performance of face recognition under varying illumination conditions.