In recent years, with the increasing demands in security monitoring, high-efficient video coding and fast browsing technique for security monitoring have received a lot attentions in industry and academy. Wherein, surveillance video flexible storage oriented scalable video coding technique and surveillance video fast browsing oriented video synopsis technique have been the hotspots in current video coding and browsing field. Through the further research on existing scalable coding and video synopsis techniques, this paper found it is neither satisfactory for distinguishing the key part with non-key part in surveillance video associated with original video synopsis technique, and nor able to perform scalable browsing on synopsis information for surveillance video. Moreover, as conventional scalable coding techniques cannot be directly compatible with video synopsis technique, it is hard to proceed scalable coding ad browsing to original surveillance video and synopsis video at the same time. Thus, in order to solve these problems, this paper has carried on a series of researches, including from the generation, browsing and coding of synopsis video in the pixel domain to the synopsis analysis directly used in the compressed domain with surveillance video. The main contribution of our work includes: (1) Aiming at the incompatible problem of conventional scalable coding techniques and video synopsis techniques, this paper has designed object flags based on quadtree labels for surveillance video: the region flag of moving objects is used to effectively label the regions that moving objects located in the procedure of video analysis; and the mapping flag of moving objects is used to quickly map moving objects in original video to synopsis video for generating synopsis video, what provides the essential intermediate condition for the following works on scalable synopsis browsing and coding. Meanwhile, an intra-frame and inter-frame coding method with object flags based on compressed-domain information is designed, it has achieved the high-efficient storage of object flags. (2) Aiming at the undistinguishable problem of key part and non-key part in surveillance video during browsing with the original video synopsis techniques, and the problem that only the generated synopsis video with fixed level can be browsed so the br...
修改评论