Spatial Neighbor Information Assisted Motion Compensated Temporal Filter for Video Coding

Motion compensated temporal filter (MCTF) is a pre-filtering technology for video encoding, which employs bilateral filtering to enhance temporal correlations among adjacent frames. In this paper, a spatial neighbor information assisted MCTF method is proposed, which introduces the spatial information into the motion estimation and bilateral filtering processes in MCTF to improve the coding efficiency. For the motion estimation process, the spatial neighbor information is involved in preventing motion estimation from capturing a local optimum. For the bilateral filtering process, the spatial neighbor information is employed as well to reduce the block boundary effect. The proposed method is implemented on top of the Fraunhofer Versatile Video Encoder (VVenC). Experimental results show that the proposed method can achieve 1.05% bitrate saving on average compared to the current MCTF scheme, and 10.33% bitrate saving compared with MCTF disabled.