Fast transform mode decision for HEVC screen content coding

Abstract

A screen content coding (SCC) standard based on high efficiency video coding (HEVC) was finalized by joint collaborative team on video coding (JCT-VC). The coding efficiency of the standard has been improved by adopting new technologies, such as intra block copy, palette mode, and adaptive motion vector resolution. However, the encoding time is significantly increased. Also, a transform skip mode is selected more frequently because traditional transform techniques for video compression cannot efficiently compress screen content videos. Therefore, we propose an early determination method for transform process in the HEVC-SCC. Residual variance is employed as a measure of complexity. Using statistics of residual variance and an online learning system, we can adaptively determine a threshold to detect the transform skip mode. Experimental results demonstrate that the proposed algorithm successfully reduces encoding time with a small coding efficiency loss.