A fusion method for robust face tracking
- 235 Downloads
Face tracking often encounters drifting problems, especially when a significant face appearance variation occurs. Many trackers suffer from the difficulty of facial feature extraction during a wide range of face turning, occlusion, and even invisibleness. In this paper, we propose a novel and efficient fusion strategy for robust face tracking. A Supervised Descent Method (SDM) and a Compressive Tracking method (CT) are employed at the same time. SDM is used to correct drifting errors of CT continuously during frontal face tracking. However, when the face orientation changes to the angle orthogonal to the view line, it results in tracking failure for the SDM method. CT is then adopted to keep the face region being tracked until SDM detects and tracks the face again. In the experiments, we test the proposed method for real-time tracking using several challenging sequences from recent literatures. The fusion strategy has achieved encouraging performance in terms of both efficiency and reliability.
KeywordsFusion algorithm Human face tracking Compressive tracking Supervised descend method
This work is partially supported by the State Scholarship Fund of China. Y. Lu is supported by the Natural Science Foundation of Heilongjiang Province of China under Grant F201428 and the 12th Five-Year-Plan in Key Science and Technology Research of agricultural bureau in Heilongjiang province of China under Grant HNK125B-04-03.
- 6.Grabner H, Leistner C, Bischof H (2008) Semi-supervised on-line boosting for robust tracking. Comput Vis Eccv 5302:234–247, Pt I, Proceedings Google Scholar
- 10.Jones MVP (2003) Fast multi-view face detection. Technical Report TR2003-96, Mitsubishi Electric Research Laboratories, Jul. 2003Google Scholar
- 11.Kaneko T, Hori O (2002) Template update criterion for template matching of image sequences. 16th International Conference on Pattern Recognition, Vol Ii, Proceedings, 1–5Google Scholar
- 12.Li HX, Shen CH, Shi QF (2011) Real-time visual tracking using compressive sensing. 2011 I.E. Conf Comput Vis Pattern Recogn (Cvpr), 1305–1312Google Scholar
- 16.Wu Y, Lim J, Yang MH (2013) Online object tracking: a benchmark. 2013 I.E. Conf Comput Vis Pattern Recogn (Cvpr), 2411–2418Google Scholar
- 17.Xiong XH, De la Torre F (2013) Supervised descent method and its applications to face alignment. 2013 I.E. Conf Comput Vis Pattern Recogn (Cvpr), 532–539Google Scholar
- 18.Xu R, Gu XD, Wei BJ (2013) An improved real time compressive tracking. 2013 Second Iapr Asian Conference on Pattern Recognition (Acpr 2013), 2013:692–696Google Scholar
- 20.Zeisl B, et al (2010) On-line semi-supervised multiple-instance boosting. 2010 I.E. Conf Comput Vis Pattern Recogn (Cvpr), 1879–1886Google Scholar
- 21.Zhang KH, Zhang L, Yang MH (2012) Real-time compressive tracking. Comput Vis Eccv 7574:864–877, Pt Iii Google Scholar
- 22.Zhang TZ et al (2012) Low-rank sparse learning for robust visual tracking. Comput Vis Eccv 7577:470–484, Pt Vi Google Scholar