A new adaptive search strategy for fast block based motion estimation algorithms

摘要

Block based motion estimation algorithms are used for video sequence compression, among which full search algorithm is the most widely used because it provides the best match in the search area but it is computationally intensive. Hence many fast block search algorithms are proposed. All fast block search algorithm (FBSA) are based on the assumption that the block distortion measure decreases monotonically as the search point moves towards the global minimum. Hence the search becomes inefficient if the complexity of the error surface increases (i.e., for multimodal or rough surface) which in turn increases the error. In this paper, we propose a new strategy which makes this FBSA adaptive for multimodal surface by taking the relative difference between the local minima and other search points and selecting more than one winner for the next step, depending upon the distortion among the search points.