Low-Resolution Face Recognition via Color Information and Regularized Coupled Mappings

Low-resolution faces are commonly encountered and significantly drop off the recognition accuracy. Color-based features are known to be less susceptible to resolution change. We use color information to deal with low-resolution face recognition. Furthermore, we propose a regularized coupled mappings which projects the face images of different resolutions into a common subspace that is most discriminative for classification. Experimental results on AR,GT and Color Feret databases show the effectiveness of the proposed method.