Fast CU size decision algorithm for VVC intra coding
Fast CU size decision algorithm for VVC intra coding
复制标题
DOI：
10.1007/s11042-023-14691-9
复制
发表时间：
2023-02-22
影响因子：
3.6
通讯作者：
Zuo, Yifan
中科院分区：
文献类型：
--
作者：
Shang, Xiwu;Li, Guoping;Zuo, Yifan
最近，为了提高编码效率，人们利用嵌套多类型树结构的四叉树建立了一种新的视频编码标准--通用视频编码(VVC)。然而，由于灵活的块划分结构，编码复杂度急剧增加。针对这一问题，提出了一种利用编码信息和纹理信息来加速帧内编码过程的快速CU大小判决算法，该算法由快速四叉树判决方法(FQD)和快速多类型树判决方法(FMD)组成。首先，通过分析相邻CU的深度相关性，可以提前预测当前CU的四叉树深度，从而跳过不必要的深度。然后，利用多类型树的率失真代价分布结合纹理信息对二叉树(BT)和三叉树(TT)的分割进行早期决策。实验结果表明，该算法在编码性能损失不大的情况下，编码复杂度平均降低了51.89%。与现有的编码方法相比，该算法在保持更好的编码性能的同时，可以获得6%~17%的额外时间节约。
Recently, an emerging video coding standard called Versatile Video Coding (VVC) is established using quadtree with nested multi-type tree structure to improve the coding efficiency. However, the coding complexity increases dramatically due to the flexible block partition structure. To tackle this issue, a fast CU size decision algorithm is proposed to speed up the intra coding process utilizing the coding information and the texture information, which consists of a fast quadtree decision method (FQD) and a fast multi-type tree decision method (FMD). Firstly, the correlation of the depths of the neighboring CUs is analyzed, which can predict the quadtree depth of the current CU early to skip unnecessary depths. Then, the distribution of the Rate-Distortion (RD) cost of the multi-type tree combined with the texture information is utilized to make an early decision on the splits of the binary-tree (BT) and ternary-tree (TT). Experimental results demonstrate that the proposed algorithm can reduce the coding complexity by 51.89% on average, with negligible loss of coding performance. Compared with state-of-the-art methods, the proposed algorithm can obtain an additional time saving from 6% to 17%, while keeps a better coding performance.