Abstract:

Video moving object segmentation is a basic problem of computer vision and video processing. In the video sequences of a dynamic scene which has global (camera) motion, accurate moving object segmentation is still a key and hot research topic. A novel video moving object segmentation algorithm based on global motion compensation and nonparametric kernel density estimation is proposed in this paper. Firstly, an efficient and accurate global motion compensation method is used to remove the motion of the background. Then the nonparametric kernel density estimation is applied to establish foreground/background probability models. Finally, the moving object can be obtained by comparing the foreground/background probability and morphological postprocessing. The experimental results demonstrate that the proposed algorithm has good results and reduces the complexity of moving object segmentation in dynamic scenes.

Key words: dynamic scene;moving object;motion compensation;kernel density estimation