The purpose of this thesis is to segment moving objects from stationary background in color video sequences. In the field of video processing, the segmentation of moving objects is a hot research topic in recent years. Varies of segmentation methods can be used according to different situations of moving objects with their different background.This paper proposed a method based on spatio-temporal information: first construct a stationary panoramic background, and then detect the moving areas in every frame. At the same time implement color image segmentation in every frame and get a series of areas which have the same color information. At last, using color information optimize the result of temporal segmentation and get accurate moving objects.In the section of temporal segmentation, construct the stationary panoramic background first. We need get every frame of the shot at first, and get the motion vectors of every frame, and calculate the motion parameters. Use the motion parameters, we can get the transformation of successive frames, and then get transformation of every two frames. We select a frame as the reference frame, construct the panoramic background. After get this background, we use every aligned frame compare with this panoramic background, so can detect which area is moving in this frame, although its not very accurate.In the section of spatial segmentation, we use a non-parametric clustering algorithm based on mean shift. At first, transfer all the pixels of color image... |