Video-based Face Recognition Using Complete Local Binary Pattern

While traditional face recognition is typicallybased on still images, face recognition from video sequenceshas become popular recently. This paper proposes a newspatiotemporal operator, CLBP-TOP, for video-based facerecognition. Traditional spatiotemporal operators for videobasedface recognition only use the sign information of localdifferences but the proposed operator uses center pixel, signand magnitude of local differences for describing andanalyzing faces in videos. The proposed method is successfullyapplied to face recognition on two standard databases,Honda/UCSD and VidTIMIT, which are provided for videobasedface recognition. The experiment results show that theproposed method significantly outperform other state-of-theartmethods in terms of recognition rate.

Keywords:

face recognition , spatiotemporal operator , local binary pattern , complete local binary pattern