Adaptive motion estimation based on spatio-temporal correlation

摘要

In this paper, we present a new motion estimation algorithm which determines an initial motion vector and a search window adaptively in video codecs which code and transmit data in a block scan order. The proposed method is based on the motion information of spatio-temporal neighboring blocks. In predicting the motion, a temporal projection technique is introduced in the determination of an initial vector, and then the search window size to refine the initial vector is determined by the correlation between neighboring blocks. With a small amount of computation during motion estimation, the proposed algorithm can estimate the motion of fast moving object. Experimental results of the proposed method show significantly improved performance in image quality and the computational complexity.