Survey on block matching algorithms for motion estimation

摘要

In video both temporal redundancies as well as spatial redundancy occurs. So to remove the both type of redundancies we need combination of systems that can remove both type of redundancies and hence we use Hybrid Video Codec for video compression. In encoder part of Hybrid Codec, we find the motion vector (MV) of the candidate block of current frame in reference frame using block based motion estimation technique. Motion estimation (ME) is to be done in the encoder side to find the best MV so that it can be applied on stored frames by motion compensated block to generate the predicted video in video compression. In this paper we have reviewed in brief already implemented various block based motion estimation techniques namely full search (FS), New Three Step Search (NTSS), three step search(TSS), diamond search algorithm (DSA), cross search(CS), hexagonal search pattern(HXSP), Adaptive Rood Pattern Search(ARPS) etc. We also try to review their performance based on the number of searching points and PSNR.