AVS2视频编码标准中熵编码模块的优化设计

Optimizations for the entropy coding module in AVS2

在对高清和超高清视频进行压缩时,编码效率不再是衡量视频压缩技术的唯一指标,为了提高视频编码器的处理速度和降低其功耗,数据吞吐率已经成为衡量视频压缩技术优劣的重要指标。作为AVS2的核心模块之一,熵编码模块在去除信源符号的统计冗余方面有着不可替代的作用。然而,在AVS2熵编码模块的设计过程中,由于没有充分考虑到数据吞吐率这一性能指标,使得其熵编码模块包含非常紧密的编码依赖关系,严重地限制着其数据吞吐率。为了解决这个问题,本文从3个方面对AVS2的熵编码模块进行了优化设计。首先,本文提出了一种快速的,与标准兼容的算术编码引擎归一化方法。该方法仅仅需要一次查表操作即可完成归一化过程。其次,本文提出了一个快速的bypass bin(概率等于0.5的二进制符号)的编解码过程,使得编解码bypass bin仅仅需要移位和加法操作即可完成。最后,本文改进了AVS2中变换系数的编码过程,来最大限度地降低变换系数之间的编码依赖关系。实验结果表明,上述3个技术可以极大地提高AVS2中的熵编码模块的数据吞吐率,同时性能损失也比较小。

When compressing HD and UHD videos, throughput should be taken into account in addition to coding efficiency for high processing speed and low power consumption. As one of the most important modules in AVS2, the entropy coding module plays an irreplaceable role on removing the statistical redundancies. However, the data throughput is not fully considered during the design of entropy coding module in AVS2, thus the strong data dependencies severely limit the improvement of throughput. To address these problems, the paper optimizes the entropy coding module in AVS2 from three aspects. First, a fast and standard-compliant normalization is proposed, in which only one look-up table is required to accomplish the normalization. Then, a fast coding process for bypass bins is proposed, in which only one addition and one shift are required. Finally, the context modeling scheme for transform coefficients is modified to reduce the coding dependencies as much as possible. Experimental results demonstrate the three techniques can significantly improve the throughput of the entropy coding module in AVS2 by keeping the similar coding performance.