A Fast Sample Adaptive Offset Algorithm for H.265/HEVC

Abstract

In this paper, a fast Sample Adaptive Offset (SAO) method is proposed to reduce the high computational complexity in SAO estimation process. On the frame level, the relationship between SAO and the quantization parameter or the layer information due to the hierarchical coding structure is explored. A fuzzy control method is proposed to decide how to perform the SAO operation. On the CTU level, an asymmetric EO/BO skipping method is proposed. The complexity measurement of a coding tree unit (CTU) is inferred by the number of coding units (CU). Edge offset (EO) or band offset (BO) is asymmetrically skipped while the CTU is of low complexity, thus the SAO process time can be reduced. Experimental results show that the proposed method can reduce about 63.89% SAO encoding time on average under the common test conditions with negligible loss compared with the original SAO method in HEVC reference software HM16.2.