Block matching algorithms in MPEG video coding

Abstract

Motion estimation is one key point in MPEG video coding allowing the use of desired bitrates in video compression. To obtain a good tradeoff between compression ratio and quality, we need more than just the reduction of the spatial redundancy in the images: the use of techniques to exploit the temporal redundancy existing in image sequences. One of such techniques is the block matching motion estimation algorithm. In this paper different search algorithms are briefly described and compared regarding matching criteria, computational complexity and search window size. Results obtained show that, depending on the type of image, there are algorithms much simpler than the full search algorithm. First results favor the adoption of algorithms designated OSA, LBS and CSA.