Improved Predictive Motion Estimation Algorithm

Motion estimation (ME) in the video encoding process affects the output quality of an encoded sequence and requires a significant part of the encoding time. Thus, low complexity motion estimation algorithms are viable solutions. In this paper, an improvement upon the predictive motion estimation algorithms is presented. The proposed method has investigated the predictor selection and the determination of threshold to terminate searching early for ME. More specifically, the inter-frame subtraction is adopted to get the motion region information of per frame. Based on the motion region information, the MV predictor sets are selected and the thresholds are determined. Moreover, the proposed method has been extended to the H.264/AVC standard. The experimental results show that the improved method can produce similar and in some cases better results compared to even the full search method while having considerably lower complexity.