Face Verification Using Color Sparse Representation

摘要

This paper proposes an effective method for face verification using color sparse representation. In the proposed method, sparse representations are separately applied to multiple color bands of face images. The complementary residuals obtained from the multiple color face images are merged by means of score-level fusion, yielding improved discrimination capability for face verification. Experimental results using two public face databases (CMU Multi-PIE and Color FERET) showed that the proposed face verification method is highly robust under challenging conditions, compared to the conventional methods using grayscale sparse representation.