一种HEVC样点自适应补偿快速算法

A fast algorithm for sample adaptive offset in HEVC

为解决高性能视频编码标准H.265/HEVC中引入的样点自适应补偿技术的计算复杂度极高导致严重影响编码效率的问题,提出一种HEVC样点自适应补偿快速算法.首先根据编码单元(coding unit,CU)的划分深度确定亮度树形编码块(coding tree block,CTB)中需要提取边缘方向的区域,然后利用边缘方向提取算法获得亮度CTB的边缘方向列表,并根据此列表减少亮度CTB在模式判别过程中遍历的模式数量,最后利用亮度和色度CTB之间的相关性进一步简化色度CTB的模式判别过程.实验结果表明,在性能损失较小的情况下,本算法可以在全I帧(AI)、随机访问(RA)、低延时B帧(LB)、低延时P帧(LP)四种配置下分别节省31.75%、56.85%、52.81%、51.51%的样点自适应补偿编码时间.

In order to solve the problem that the high computational complexity of the adaptive compensation technique for samples introduced in the high performance video coding standard H.265/HEVC seriously affects the coding efficiency,we propose a fast SAO algorithm for HEVC.First,according to the partition depth of the coding unit(CU),determine the area in the luminance coding tree block(CTB)that needs to extract the edge direction,and then use the edge direction extraction algorithm to obtain the edge direction list of the luminance CTB,according to this list,the number of modes that the luminance CTB traverses in the mode discrimination process is reduced,finally,the correlation between the luminance and the chrominance CTB is used to further simplify the mode discrimination process of the chrominance CTB.Experimental results show that the proposed algorithm can save 31.75%,56.85%,52.81%and 51.51%of the SAO coding time under the four configuration conditions of all intra(AI),random access(RA),low delay with B pictures(LB),and low delay with P pictures(LP)respectively,while only causing negligible performance loss.