An Ultralow Complexity String Matching Approach to Screen Content Coding in AVS3
An Ultralow Complexity String Matching Approach to Screen Content Coding in AVS3
复制标题
AVS3 中屏幕内容编码的超低复杂度字符串匹配方法
DOI：
10.1109/tcsvt.2020.3029726
复制
发表时间：
2021-09
影响因子：
8.4
通讯作者：
Lin Tao
中科院分区：
文献类型：
--
作者：
Yang Yufen;Zhou Kailun;Zhao Liping;Lin Tao
屏幕内容编码(SCC)在移动设备中的应用越来越广泛，为了降低功耗，需要超低的编码复杂度。本文针对SCC提出了一种超低复杂度的字符串匹配方法。该方法有两个基本特点：1)只允许两种最有效的参考字符串，它们具有两个最频繁出现的偏移量(即，位移向量)，称为对偶单位偏移量；2)充分优化字符串匹配参数的编码，以获得最大的编码效率。由于对于任何正在编码的当前串，最多只允许两个偏移值和对应的参考串位置，所以最优参考串搜索过程非常简单，并且只需要从最多两个候选中选择最佳串。此外，在该方法中，字符串匹配参数，即字符串偏移量和字符串长度的取值范围非常有限，导致编码这两个参数的位数很少，编码效率很高。与禁用IBC的AVS3参考软件HPM7.0相比，使用AVS3 SCC通用测试条件和带有运动类别的文本和图形的YUV测试序列，在超低的编解码复杂度下，对于所有帧内(AI)和低延迟B(LDB)配置，所提出的技术分别获得了10.8%和5.8%的Y平均BD降低。
Screen content coding (SCC) is increasingly used in mobile devices, and ultralow coding complexity is required for low power consumption. This article proposes an ultralow complexity string matching approach to SCC. The proposed approach has two essential features: 1) allowing only two types of most effective reference strings that have two most frequently occurring values of offset (i.e., displacement vector) named dual unity offset and 2) fully optimized coding of string matching parameters for maximum coding efficiency. Since at most only two offset values and corresponding reference string positions are allowed for any current string being coded, the optimal reference string searching process is extremely simple and only needs to select the best string from at most two candidates. Moreover, in the proposed approach, the value ranges of string matching parameters, i.e., the string offset vector and string length are very limited, resulting in few bits to code the two parameters and high coding efficiency. Compared with the AVS3 reference software HPM7.0 with IBC disabled, using AVS3 SCC common test condition and YUV test sequences in text and graphics with motion category, the proposed technique achieves Y average BD-rate reduction of 10.8% and 5.8% for all intra (AI) and low-delay B (LDB) configurations, respectively, at ultralow encoding and decoding complexity.