Implementation of Face Detection System Using Future Extraction

Abstract

Face detection and clustering is more significant in the Image sequence because human faces have unlimited orientations and positions. This paper proposed a Spatio Temporal method for clustering human faces from the image sequence. The proposed method has three main stages. Initially a face detector is used to localize faces in all frames of the video and extract several features of the detections. Viola jones detector and SURF detector both are used for face detection and feature extraction respectively. Optical flow estimation is utilized for computing the movement among any of two frames. Pyramidal version of lucas kanade algorithm is used in optical flow estimation. This estimation is used to determine a dissimilarity matrix like appearance dissimilarity and time dissimilarity. Clustering tracks the same faces from the image sequence but it may have different pose, image orientation. Finally, an optimization method involved in clustering of faces.