改进的基于AVS2的去块效应滤波算法  

Improved de-blocking filter algorithm based on audio video coding standard

作　　者：陈智贤 王国中[1] 赵海武[1] 李国平[1] 滕国伟[1] 

机构地区：[1]上海大学

出　　处：《电视技术》2017年第9期6-13,26,共9页Video Engineering

基　　金：国家高技术研究发展计划(863计划)项目(2015AA015903);国家自然科学基金项目(61271212)

摘　　要：在H.265/AVS视频编码标准中,去块效应滤波是环路滤波模块最先进行的,对提高整体编码效果起至关重要的作用。针对现有去块效应滤波算法对提高图像质量方面的不足,提出一种改进的基于像素梯度和色度分量预测、变换单元划分方式的滤波算法。对于亮度块,该算法主要通过建立亮度平坦度阈值的回归模型,根据不同图像的纹理复杂度,为每帧图像自适应地选择不同的阈值,从而更精确地判断出每条边界的滤波强度。对于色度块,采用与亮度不同的滤波块尺寸及滤波强度的判断方式,在不增加色度滤波复杂度的同时,提高色度分量的滤波质量。实验结果表明,与AVS2标准下的滤波计算结果相比,改进的算法不仅提高了6%的亮度和24%的色度去块效应滤波性能,而且减少了24%的编码时间。De-blocking filter is not only first performed in the part of in-loop filter according to video coding standard H. 265 / AVS, but also plays an important role in the increase of entire encoding quality. Aiming at the problem that existing de-blocking algorithms still have deficiencies on improving the quality of coding image, an improved filter algorithm based on pixe gradient, the division of prediction and transform chroma units is proposed in this paper. In order to determine the filter strength of every bound- ary more accurately, a regression model of flatness threshold of luma component is established to adaptively select different thresh- old for each image by computing its texture complexity. For the chroma component, by using different size of filter blocks and ways of boundary strength judgment, the filtering quality of chroma is improved without increasing the complexity. The experimental re- sults show that, compared with the calculated results in AVS2 standard, the proposed algorithms not only improves the effect of de -blocking algorithms for luma coomponent by 6% as well as 24% for chroma, but also reduces encoding time by24%.

关 键 词：H.265/AVS 去块效应滤波(de-blockingfilter) 自适应选择阈值 色度滤波强度判断 提高整体编码质量 

分 类 号：TN919.81[电子电信—通信与信息系统]