基于文字块识别和长方形哈希块匹配的屏幕内容编码优化方案

发布时间：2018-09-06 08:01

【摘要】：高效视频编码(HEVC,High Efficiency Video Coding)作为当前最新的视频编码标准,在相同视频质量水平下可比H.264提高大约一倍的压缩率,并提供对高清、超高清分辨率视频的支持。屏幕内容编码(SCC,The Screen Content Coding)作为HEVC的重要扩展标准,被广泛应用于远程桌面协助、实时视频共享、以及无线显示等领域。SCC在HEVC的基础上,针对屏幕内容图像特性增加了五种编码工具,提升了屏幕内容编码的性能。然而,由于屏幕内容编码是一个新兴的课题,现有标准还存在编码复杂度高、编码效率有待进一步优化等问题。对于SCC编码时间复杂度高的问题,论文针对带有大量文本内容的视频序列展开研究,提出了一种基于文字块识别的编码优化算法。通过分析SCC中两个编码工具(帧内块拷贝、帧内方向预测)的编码特点,将当前编码块划分为文本块和非文本块,并分别采用不同的处理策略,降低了编码的时间复杂度。实验结果表明,在all intra配置下,测试序列的BD-rate只增加了 0.44%,编码时间平均减少了 23%。哈希块匹配作为屏幕内容编码工具之一,在不改变解码框架的情况下,对带有大量文字的视频能够显著提高编码效率。但由于哈希块匹配只使用正方形的块进行预测,预测精度有待提高。论文在分析哈希块匹配过程的空间和时间复杂度之后,设计了 8×4和4×8两种哈希块匹配过程,能够在整帧范围内搜索到更精确的匹配块,提高了编码效率。实验表明,在low delay和random access两种配置下,全测试序列的BD-rate平均减少了 0.6%和0.3%,而编码时间仅仅增加了 1.56%和 1.23%。
[Abstract]:Efficient video coding (HEVC,High Efficiency Video Coding), as the latest video coding standard, increases the compression ratio by about twice as much as H. 264 at the same video quality level, and provides support for high-definition, ultra-high-definition video. Screen content coding (SCC,The Screen Content Coding), as an important extension standard of HEVC, is widely used in the fields of remote Desktop Assistance, Real-Time Video sharing, and Wireless display based on HEVC. To improve the performance of screen content coding, five kinds of coding tools are added to screen content image characteristics. However, because screen content coding is a new topic, there are still some problems such as high coding complexity and high coding efficiency. For the problem of high time complexity of SCC coding, this paper presents an optimized coding algorithm based on text block recognition for video sequences with a large amount of text content. By analyzing the coding characteristics of two coding tools in SCC (intra-block copy, intra direction prediction), the current coding block is divided into text block and non-text block, and different processing strategies are adopted to reduce the time complexity of coding. The experimental results show that under all intra configuration, the BD-rate of the test sequence is only increased by 0.44%, and the coding time is reduced by 23% on average. As one of the screen content coding tools, hash block matching can significantly improve the coding efficiency of video with a large amount of text without changing the decoding framework. However, because the hash block matching only uses square blocks to predict, the prediction accuracy needs to be improved. After analyzing the space and time complexity of the hash block matching process, two kinds of hash block matching processes, 8 脳 4 and 4 脳 8, are designed, which can search for more accurate matching blocks in the whole frame, and improve the coding efficiency. The experimental results show that the BD-rate of the whole test sequence decreases by 0.6% and 0.3% on average in both low delay and random access configurations, while the coding time increases only by 1.56% and 1.23%.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【相似文献】