Multiple face location using motion information

Face location is a difficult problem for face recognition and multiple face location is more challenging. In this paper, two new methods are presented for multiple face location via motion analysis techniques. The first method is based on motion segmentation. The authors introduce a new segmentation method by computing optical flow only on the Motion Zero-Crossing Boundary (MZCB) followed by a simple clustering method to segment each person. Then an intuitive but effective location algorithm is applied to locate each face. The second method is derived from the Hough Transform (HT). After modeling a head outline as a curve consisting of circle segments, a modified HT is used to find the center of each face. Finally, the two methods are compared and the future research directions are given.