Prediction-Based Asymmetric Pattern Search Algorithm for Fast Block-Matching Motion Estimation

In this paper, we proposed a novel fast block-matching algorithm for motion estimation, called prediction-based asymmetric pattern search (PAPS). The initial pattern of the PAPS is a very compact center-biased pattern, which has smallest search points. The PAPS utilizes the block distortions of search patterns to decide the search direction, and then an asymmetric search patterns are applied to the subsequent steps accordingly. In new method, the search points are greatly reduced to locate a motion vector. Furthermore, a prediction scheme is incorporated to favor larger motion blocks. Experimental results show that the proposed method has the fastest search speed among all algorithms in the simulation for all test sequences with very close PSNR value.