Image synthesis for sparse representation based multi-pose face recognition

In this paper we consider the problem of the pose variance in face recognition. In practice, the face images we get usually have different poses; however the number of training sample is limited, so the sparse representation based classification (SRC) for face recognition cannot work well. For the purpose of expending the sample size, a 3D face modal based image synthesis method is adopted. Experiments on the CMU-PIE face are conducted to verify the efficiency of the proposed image synthesis based SRC (IS-SRC) method for multi-pose face recognition.