Implementation, Analysis And Improvement Of Motion Estimation Algorithms

Motion estimation, which may refer to image-plane motion (2-D motion) or object-motion (3-D motion) estimation, is one of the fundamental problems in digital video processing. Indeed, it has been the subject of much research effort. Block-base motion estimation and compensation are among the most popular approaches.Block-base motion estimation has been adopted in the international standards for digital video compression, such as H.261 and MPEG 1-2. Block-based motion estimation is also widely used in several other digital video applications, including motion-compensated filtering for standards conversion.In this thesis, an attempt is made to implement and improve the performance of motion estimation algorithms. First, Block-based full search, which evaluates the matching criterion at each pixel, is implemented. Then the impact of matching criterion function to the correctness and performance of the algorithms is discussed and analyzed. Also, the impact of the size of block and window is discussed. Later, the three-step search algorithm, which is to estimate motion vectors on a sparse grid of pixels, is implemented. It is an example of faster search algorithms. Some improvements to the algorithm are made, which include special processing of pixels on the edge of the image, adjustment of the matching criterion functions. Then the performance (including veracity and time cost) of these two algorithms is compared.