Sample Adaptive Offset Optimization in HEVC

As the next generation of video coding standard, High Efficiency Video Coding (HEVC) adopted many useful tools to improve coding efficiency. Sample Adaptive Offset (SAO), is a technique to reduce sample distortion by providing offsets to pixels in in-loop filter. In SAO, pixels in LCU are classified into several categories, then categories and offsets are given based on Rate-Distortion Optimization (RDO) of reconstructed pixels in a Largest Coding Unit (LCU). Pixels in a LCU are operated by the same SAO process, however, transform and inverse transform makes the distortion of pixels in Transform Unit (TU) edge larger than the distortion inside TU even after deblocking filtering (DF) and SAO. And the categories of SAO can also be refined, since it is not proper for many cases. This paper proposed a TU edge offset mode and a category refinement for SAO in HEVC. Experimental results shows that those two kinds of optimization gets -0.13 and -0.2 gain respectively compared with the SAO in HEVC. The proposed algorithm which using the two kinds of optimization gets -0.23 gain on BD-rate compared with the SAO in HEVC which is a 47 % increase with nearly no increase on coding time. Copyright © 2014 IFSA Publishing, S. L.