Research On Change Detection Video Segmentation Method

Content-based processing is the future way of multimedia processing. Actually, it's the higher stage of information processing and closer to the manner that human use to process information.MPEG-4 and the latest MPEG-7, MPEG-21 are the new generation video coding standards designed under such development trend. The segmentation and generation of video objects is the key and basis of content-based coding and interactive operations, and it has direct impact on the performance of video coder and validity of video representation. Video segmentation is a very important supporting technology in the future video coding area, so research on video segmentation has great significance and application value.In this thesis, we classify the existed video segmentation algorithms, compare and analyze these algorithms in different aspects. We focus on the research of change detection and Spatial-temporal segmentation. Three different segmentation solutions are proposed according to different application scenarios, and several adjustments are made in the essential modules: to get the adaptive threshold, Least-Half Sample method is used to estimate the mean and variance of the background noise; Morphological gradient operator is used to enhance the single-pixel edge acquired by Canny edge detector. Finally, based on the requirement analysis of E03(scan wall) in Guangdong Exhibition hall, we propose an efficient segmentation method for the application system.The testing results demonstrate the efficiency of our solutions in different application scenarios, show that our research has some significance in theory and in application and establish a solid basis for future research on video segmentation.