A Novel Automatic Motion Segmentation Method Based on Optical Flow

Motion segmentation is one of the fundamental problems in digital image processing area. The estimation of velocity field is one of the most notable approaches on the first step of motion segmentation. A novel motion segmentation method is proposed, which is nonparametric and automatically selects value of the segmentation criterion. In particular, we first employ the optical flow to estimate the motion field; then we establish a segmentation criterion and measure the difference between regions of the moving objects and background separated in the segmentation criterion at each level; next we adopt the level of the segmentation criterion which makes the two regions the most different; finally, our approach output regions of moving objects and background from the scene. Experimental results demonstrate that our method is effective, and the procedure is much simpler than the methods often used.