- Book Chapter
4
- 10.1007/978-3-030-03000-1_3
Advances and Trends in Video Face Alignment
- Dec 15, 2018
- Gang Zhang + 3 more +3
Face alignment in a video is an important research area in computer vision and can provides strong support for video face recognition, face animation, etc. It is different from face alignment in a single image where each face is regarded as an independent individual. For the latter, lack of amount of information makes the face alignment an under-determined problem although good results have been obtained by using prior information and auxiliary models. For the former, temporal and spatial relations are among faces in a video. These relations can impose constraints among multiple face images each other and help to improve alignment performance. In the chapter, definition of face alignment in a video and its significance are described. Methods for face alignment in a video are divided into three kinds: face alignment using image alignment algorithms, joint alignment of face images, and face alignment using temporal and spatial continuities. The first kind of face alignment is studied and some of surveys have described the work. The chapter will mainly focus on joint face alignment and face alignment using temporal and spatial continuities. Herein, some representative methods are described, and some factors influencing alignment performance are analyzed. Then the state-of-the-art methods are described and the future trends of face alignment in a video are discussed.
Read more