摘要

A new scene segmentation method based on the distribution of motion intensity was proposed in this paper. Firstly, the motion foreground was extracted by fusing time and spatial information. Furthermore, noise was eliminated after motion vectors were extracted via pyramid-style Lucas-Kanade optical flow method. Finally, a frame scene was divided into blocks by proposed pyramid-style division method and the intensity mask generated by computing the motion intensity for each block was used to scene segmentation. The experimental results show that the motion intensity measurement is easy to compute and it has an outstanding performance on region segmentation and scene discrimination.