AVS解码算法研究及其在DSP上优化实现

发布时间：2018-05-20 00:29

本文选题：AVS解码器 + DSP　；参考：《太原理工大学》2013年硕士论文

【摘要】：AVS(数字音视频编解码技术标准)作为我国自主知识产权的信源编码标准,其视频部分标准具有压缩编码效率高、算法实现复杂度低的特点,适于嵌入式平台上开发,能广泛应用于数字电视、网络摄像机等领域。随着嵌入式系统的飞速发展,使用体积更小,外观更精美的移动终端传输视频流日益成为主流,德州仪器公司提供的开放式多媒体应用平台——OMAP3530(Open Multimedia Application Platform)是一种片上系统(SoC),不仅体积和功耗很小,更能高效地处理多媒体数据视频流。针对在OMAP3530平台上实现AVS视频解码器实时解码704×576格式视频流,本课题主要研究内容、所做工作和取得的成果如下：本文首先介绍了AVS视频标准系统结构,分析了解码器整体框图,从解析和解码两部分讨论了AVS视频解码器相关核心技术,着重介绍了熵解码,帧内、帧间预测和滤波等模块。然后针对DSP的特点,本文调整了AVS视频解码器整体流程,以便移植到DSP平台,将Ⅰ帧和P、B帧解码分开,给出解码器重要模块实现流程图,并通过DSP平台测试工具对解码器主要函数进行解码复杂度分析。最后本文在C和汇编级别上优化了解码器核心模块。重点分析了AVS熵解码可变长码解码特点,通过不同的分组对码字重新排序来改进查表方式,降低了DSP实现的复杂度。对码表存储结构做相应的调整并通过汇编语言实现,降低了查表复杂度。为进一步提高解码速度和图像质量,本文还提出一种新的边界环路滤波模式,最大程度的重复利用当前宏块的数据,并通过汇编代码实现,节省了环路滤波所需的时间。经综合测试表明,在不改变图像质量的前提下,采用汇编指令优化程序,不仅节省了大量的存储空间,而且代码执行周期仅为原解码器的23.8%,极大的提高了解码效率,达到实时解码的要求。
[Abstract]:AVS (Digital Audio Video coding and Decoding Technology Standard) is the source coding standard of our country's independent intellectual property rights. The video part of AVS standard has the characteristics of high compression coding efficiency and low algorithm complexity, so it is suitable for the development of embedded platform. Can be widely used in digital television, network cameras and other fields. With the rapid development of embedded systems, the use of smaller, more beautiful mobile terminals to transmit video streams has become the mainstream. OMAP3530Open Multimedia Application Platform), an open multimedia application platform provided by Texas Instruments, is a kind of on-chip system, which not only has small volume and power consumption, but also can process multimedia data video stream efficiently. In order to realize the real-time decoding of 704 脳 576 format video stream by AVS video decoder on OMAP3530 platform, the main research contents of this paper are as follows: In this paper, the architecture of AVS video standard system is introduced, the whole block diagram of decoder is analyzed, the core technology of AVS video decoder is discussed from the two parts of analysis and decoding, and the modules of entropy decoding, intra frame, inter frame prediction and filtering are introduced emphatically. Then, according to the characteristics of DSP, this paper adjusts the whole flow of AVS video decoder in order to transplant it to DSP platform, separates frame I and B frame, and gives the flow chart of the important module of decoder. The decoding complexity of the main functions of the decoder is analyzed by using the DSP platform test tool. Finally, the core modules of the decoder are optimized at C and assembly levels. The characteristics of variable length decoding based on AVS entropy decoding are analyzed emphatically. The table lookup method is improved by reordering different block codewords, and the complexity of DSP implementation is reduced. The code table storage structure is adjusted and realized by assembly language, which reduces the complexity of table searching. In order to further improve the decoding speed and image quality, this paper also proposes a new edge loop filter mode, which can reuse the current macroblock data to the maximum extent, and realize it by assembly code, which saves the time needed for loop filtering. The comprehensive test shows that, without changing the image quality, using the assembly instruction optimization program not only saves a large amount of storage space, but also the code execution period is only 23.8 of the original decoder, which greatly improves the decoding efficiency. Meet the requirement of real-time decoding.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TN919.81

【参考文献】