Multi-dimensional Vector Matrix Adaptive Block Coding Based On Object-tracking

Video for its vivid, intuitive and informative, has become major source of knowledge for human understanding the world. However, the huge amount of data to the computer storage and network transmission made great difficulties, which blocked human to acquire and use of video and image. Video compression technology reduced the amount of video data effectively, and played an important role on the storage and transmission of video. At present, a common feature of the video compression is that no matter what the specific content of the image is, for algorithm simple, the fixed block size is carried out. This has ignored the characteristics of the image to some extent.Because the final recipient of video and image is human, according to the characteristics of human vision, people paid more attention to specific parts of the video, that is the non-smooth region mainly located in the center of the image, and hoped it had high resolution, however, for the left parts of the image to reach the basic visual requirements.Therefore, applied low compression ratio or the lossless compression to non-smooth region which located in the image center, and high compression ratio to smooth area or less important area. The focus of this paper is to use Canny edge detection operator to extract the edge of the video target, and then to extract the foreground region, that is region of interest, using object tracking. Under the premise of guaranteeing the quality of foreground region, improve the coding efficiency and compression quality, the paper blocked the foreground region with smaller size, coded it preferentially or with lower compression ratio, and the left of the video or image with larger size, coded it with higher compression ratio.The paper combined adaptive blocking based on target tracking algorithm with multi-dimensional vector matrix, and programmed multi-dimensional vector matrix adaptive block codec based on object tracking for video and image with Visual C++ 6.0. Compared the experiment results with fixed block coding algorithm, it can been see that the method is feasible and effective.