A Real-time Video Object Segmentation Algorithm Based on Change Detection and Background Updating

This paper proposes an efficient automatic video object segmentation algorithm by combining change detection with background updating. First, motion information is obtained from difference between previous frame and current frame instead of complicated motion estimation. Second, background subtraction is introduced to solve the problems of still object and uncovered background an ill-posed in conventional method. Moreover, the effect of shadow and residual background can be suppressed by a novel hierarchical segmentation in boundary refinement step. Finally, experimental results show that the proposed algorithm can maintain a high accuracy upper 95 percent with capturing situations of the fixed camera.