Fast sub-pixel motion estimation based on directional information and adaptive block classification

Abstract

Sub-pixel motion estimation takes a significant amount of time in motion estimation of the H.264 encoder. A few fast schemes have been proposed. Most of them are based on the analysis of the neighboring blocks. In this paper, we propose a new approach of fast sub-pixel motion estimation. This algorithm uses the statistical information, which features the motion activities of the blocks in the previous frame, to predict the characteristics of the motion activities of the blocks in the current frame. Both intermediate results of sub-pixel motion estimation of the current block and motion vectors of Macro Blocks (MB) in the previous frame are considered. With this prior knowledge, we can skip some of the candidate checking points and make a decision on early termination before quarter-pixel motion estimation. Extensive experimental work has been done, results of which show that our approach could save 83.64% of the computational time over that of the Full Position sub-pixel motion estimation algorithm in JM, with a negligible degradation in peak signal-to-noise ratio (PSNR)and Bits performance.