Adaptive motion estimation in video coding with a stochastic model

A new motion vector (MV) estimation method for video image compression based on a stochastic model is presented in this work. It is well known that the location of the smallest sum of absolute difference (SAD) does not always give the true MV since the MV obtained via full block search is often corrupted by noise. Thus, multiple locations of relatively small SAD are searched with an adaptive search window by using our method. We pick an MV among those candidates by using temporal correlation. Furthermore, since temporal correlation reveals the noise level in a particular region of the video image sequence, we are able to reduce the search area very effectively. The excellent performance of the proposed method is demonstrated by numerical experiments