Using concentric circular invariant features for face image retrieval

An effective face image retrieval method based on concentric circular invariant features (CCIF) is presented in this paper. With this method, each face image is firstly segmented into several concentric circular subimages based on the square-to-circular image transformation. The local invariant features robust to image translation and rotation for each subimage are then extracted in the new polar coordinates space. The experimental results on AR gray face database show that the proposed features are robust to different facial factors of expression, pose, illumination and partial occlusion, and can get better performance of retrieval precision and speed simultaneously.