A Face Superresolution Algorithm in Video Sequence of Monitoring System Based on Patch Reconstruction

The faces in video sequence of monitoring system are the key information for crime detection, identity recognition and other fields. However, the face is usually blur for the far distance or low-resolution of camera. In this paper, a face super resolution algorithm based on patch reconstruction was presented to enlarge and make face image clear. The face images were segmented from the sequence by frame-difference method at first. Then, Patches were constructed in the middle frame of the sequence and matched with other frames. The matched patch that had minimum differences in one of other frames was used to reconstruct the high-resolution face image. Experimental results showed that the proposed algorithm provided clearer and sharper details compared to the existing algorithms.