Temporal correlation based general search strategy for multiple reference frame motion estimation

Abstract

Multiple reference frame motion estimation (MRFME) is one of the main features of recent video coding standards such as H.264/AVC and high efficiency video coding. This paper presents a fast search strategy for avoiding redundant computations during MRFME. The proposed method reduces the number of pixels to be examined using the precalculated rate-distortion costs of previously searched frames. Experimental results show that the proposed method reduces the complexity of the motion estimation (ME) process by up to 81.43 %.