Motion Estimation Using Cross Center-Biased Distribution and Spatio-Temporal Correlation of Motion Vector

Abstract

Motion estimation (ME) technique is an important part of video encoding, since it could significantly affect the compression ratio and the output quality. But full search (FS) for block-based motion estimation is computationally intensive and it can hardly be applied to any real time applications. In this paper, we propose a new adaptive motion vector estimation algorithm based on the cross center-biased distribution property and the high spatio-temporal correlation of motion vector to find the motion vector more efficiently. The proposed method identifies a more accurate first search center instead of using the origin as the initial search center through compensating the search area based on the spatio-temporal correlation of motion vector and defines a search pattern adaptively. As a result, we reduce the total number of search points used to find the motion vector of the current block and improve the motion estimation accuracy. Experiments show that the speedup improvement of the proposed algorithm over Diamond Search algorithm (DS), Motion Vector Field Adaptive Search Technique (MVFAST) and Predictive Motion Vector Field Adaptive Search Technique (PMVFAST) can be up to 1.3 ~ 2.8 times on average and the image quality improvement can be better up to 0.1(dB)~ 0.2(dB) compare with PMVFAST.