Content Awareness-Based Fast Block Motion Estimation Algorithm for H.264/AVC Encoding System

Motion estimation part in video compression has high computational complexity. Still a lot of researches for fast scheme are in progress. In this paper, we propose an efficient motion search method based on content-awareness. The proposed method consists of three steps. The first thing is the motion classification for using the center position distribution. The second step is the calculation of predictors according to motion classification. The last is to set the search pattern size by using between predictor and center position distance, adaptively. Due to not utilization of temporal-spatial information, the proposed algorithm can reduce the required memory requirement. Experimental results show that the proposed algorithm can achieve speed-up factor of up to 30.66% on average with quite good bit rate performance, comparing to UMHexagonS and full search method and based on the H.264/AVC reference software (JM).