A Still-to-Video Face Verification System Using Advanced Correlation Filters

In this paper we propose to use the technology of multiple correlation filters to address the issues involved in still-reference-to-video-sequence face verification. In particular, we assume that the gallery consists of one or a few digital images of a person's face as reference and the probe set consists of video sequences as test data. The test data consists of a sequence of face images of a naturally moving person, captured from one or more cameras. An automatic face recognition system involves face/head tracking and cropping, correlation filter based face matching, and combining the evidence from multiple frames of the face images.