Abstract

Face recognition under viewpoint and illuminationchanges is a difficult problem, so many researchers havetried to solve this problem by producing the poseandilluminationinvariant feature. Zhu et al. [26] changed allarbitrary pose and illumination images to the frontal viewimage to use for the invariant feature. In this scheme, pre-serving identity while rotating pose image is a crucial is-sue. This paper proposes a new deep architecture basedon a novel type of multitask learning, which can achievesuperior performance in rotating to a target-pose face im-age from an arbitrary pose and illumination image whilepreserving identity. The target pose can be controlled bythe user’s intention. This novel type of multi-task model significantly improves identity preservation over the single task model. By using all the synthesized controlled pose images, called Controlled Pose Image (CPI), for the poseilluminationinvariant feature and voting among the multiple face recognition results, we clearly outperform the state-of-the-art algorithms by more than 4?6% on the MultiPIE dataset.