One-dimensional processing for efficient optimal post-process/in-loop filtering in video coding

Abstract

In this paper, efficient optimal post-process/inloop filtering in video coding is presented by applying two separate one-dimensional filters to significantly decrease the computational load of the optimal filtering process at the expense of only a slight quality loss. Furthermore, the compression of filter coefficients is evaluated using a Golomb-Rice (GR) coding approach together with differential encoding. Experimental results show that the proposed one dimensional processing approach reduces the computational complexity by at least 4 times compared to the formerly proposed two-dimensional optimal filtering case and enables real-time implementation. It is furthermore shown that, the filter coefficients can be compressed by up to 40% if GR coding is used together with differential encoding.