Structural Similarity based Image Quality Map for Face Recognition across Plastic Surgery

Variations in the face appearance caused by plastic surgery on skin texture and geometric structure, can impair the performance of most current face recognition systems. In this work, we proposed to use the Structural Similarity (SSIM) quality map to detect and model variations due to plastic surgeries. In the proposed framework, a S-SIM index weighted multi-patch fusion scheme is developed, where different weights are provided to different patches in accordance with the degree to which each patch may be altered by surgeries. An important feature of the proposed approach, also achieving performance comparable with the current state-of-the-art, is that neither training process is needed nor any background information from other datasets is required. Extensive experiments conducted on a plastic surgery face database demonstrate the potential of SSIM map for matching face images after surgeries.