Optimization Of Loop-filtering Algorithms On HEVC Video Coding Standard

With the increasing requirements of high- level video experience, the process of super/high definition videos has become the main development trend of the modern video techniques. However high definition video contents are becoming big challenges for transferring and storaging technology. In order to achieve higher compression efficiency, more and more coding techniques such as enhanced loop-filtering technique, have been added to the latest video coding standard HEVC which follows H.264’s hybrid coding framework and the general ideas.Loop-filtering comprises mainly of deblocking filter and sample adaptive offset, which optimize the blocking artifacts in reconstruction images, and efficiently reduce the compression distortion caused by prediction and quantization. Furthermore, cumulative image distortion can be reduced by using the reconstruction images as reference pictures which are more precise to original images. However, the comparison of corresponding video sequences in HM12.0 software shows that the coding efficiency of HEVC can not be guaranteed for video sequences of 4:4:4 format or bit-depth larger than 10. This paper focus on the loop- filtering technology of HEVC, and presents two kinds of improved methods. The main contributions contains the following two aspects:Firstly, we propose an improved algorithm for 4:4:4 format chroma deblocking filter. In HEVC RExt, chroma components in 4:4:4 format can contain much more information compared to those in 4:2:0 format. This implies that the blocking artifacts in chroma components can degrade the overall visual quality. Two methods are proposed to solve the problem. In the first method, based on the boundary strength of filter chroma deblocking filter is applied. In the second method, the chroma filter is used when the boundary strength is larger than 0. It is also proposed to signal deblocking control parameters separately for chroma components. Experimental results report from common JCT-VC test sequences indicates visual quality improvements as well as coding efficiency improvements of 0.2/2.3/2.2 and 0.4/2.1/1.9(Y/U/V BD-rate(%)) for RA and LB Main- tier in Method 1, and 0.1/0.5/0.6 and 0.0/0.4/0.5 in Method 2, respectively.Secondly, we propose an improved algorithm to apply adpative offset scaling for luma and chroma separately with independent scale factors when bit-depth is larger than 10. The algorithm considers not only offsets with larger magnitude which are still necessary, but offsets with lower magnitude which are more frequently used. Experimental results show coding efficiency improvements by 0.3/0.2/0.2(%) for Y/Cb/Cr in case of All-Intra-HE-Main-Tier with internal bit-depth increase by 2, and 0.3/0.4/0.4(%) with internal bit-depth increase by 4. The scaling parameters can be packed in the slice header so that offset scaling can be adaptive to input sequence characteristics.