96 Condensation Method for Face Tracking Based on Features Fusion
Download citation file:
- Ris (Zotero)
- Reference Manager
A robust face tracking algorithm is proposed to color video sequence of complex background, in which uses a fusion of gradient model and color histogram model as clues. First, the method constructs histogram model according to color characteristics of high clustering about H-channel of HSV, Cb-channel and Cr-channel of YCbCr, and a gradient model according to gradient characteristics in the elliptic border is also constructed. Second, an improved CONDENSATION filtering algorithm is applied to face tracking, in which the problem of occlusion is solved better in video sequence. Experiment result shows the robustness of the method against the complex background of face rotation, partial occlusion and skin color distraction.