节点文献
基于DM642的视频图象压缩系统及其关键技术研究
Realization of DM642-Based Hardware System and Research on Key Techniques of Video Compression
【作者】 李鹏；
【导师】 张钧萍；
【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2007， 硕士
【摘要】 电视制导是现代战争中应用广泛的一种精确制导方式,它是以视频信号实时传输为基础的。数字电视制导有效解决了传统模拟制导系统占用带宽宽、易受干扰、硬件复杂、体积大、不易标准化等问题,然而在实时传输中数字视频信号的高数据率与有限的传输带宽之间产生了矛盾,需要借助视频压缩技术来解决。因此,研究高保真、大压缩比、实时性强的视频压缩系统具有重要的理论和应用价值,是数字电视制导系统中亟待解决的课题。传统的视频压缩方法Motion-JPEG压缩效率不高、MPEG延迟较大, H.264作为新一代视频压缩编码国际标准是一种高效、低延迟的压缩编码方法,更加适合电视制导的应用。为此,本文重点对H.264压缩标准进行研究,并在此基础上设计实现电视制导图象压缩系统。论文首先总结了多种图象视频编码标准的技术框架,对各种标准进行了比较研究,尤其是对H.264编码标准的基本框架及基本原理进行了分析,重点研究了帧内、帧间预测模式,整数离散余弦变换与量化,熵编码等模块中的关键技术,分析了H.264所具有的独特优势,为本课题硬件设计与实现打下了坚实的理论基础。德州仪器生产的TMS320DM642数字信号处理器DSP专门设计用来处理视频多媒体信号,具有主频高、片上资源丰富等特点。本课题就是基于DM642处理器设计并实现双核并行处理的高速实时视频压缩处理平台。该平台通过DSP和FPGA双核配合工作,实现图象的采集、压缩、传输、解压缩、显示等功能;同时该平台留有以太网等扩展接口,可以实现目标跟踪等其他算法,具有很强的可扩展性和通用性。系统最终实现每秒压缩CIF格式图象20帧,基本实现实时压缩;在20倍以上压缩比的条件下,取得了33.72dB的峰值信噪比,获得了良好的图象恢复效果。快速的帧内、帧间预测算法是图象压缩系统中的关键技术,本文对已有的预测算法的原理和组成模块进行了深入的分析,在此基础上根据课题的需要,研究了三种基于搜索模式的优化运动估计算法,并且针对不同信道带宽和不同类型的视频序列进行了压缩仿真,实验表明改进后的算法对提高动态视频图象的压缩效率和恢复质量具有良好的效果。
【Abstract】 TV weapon guidance is one of the widely used accurate weapon guidance modes in the modern war, which is based on the real time transmission of the video signal. Analog video signal is used by the conventional TV guidance system during the transmission, which needs wide bandwidth and is prone to interference. Furthermore, the hardware of the system is complicated and voluminous, which cannot easily be standardized. Employing digital TV guidance system is a way to solve this problem. However, there exists a great conflit between high data rate of the digital video signal and limited bandwidth of the channel during the real time transmission, which can be solved by video compression technique. Therefore, the research of the video compression system with high fidelity, large compression ratio, low delay is of great theoretical significance and application value, and is also an urgent task in the research and development of the digital TV guidance system.In the conventional video compression methods, Motion-JPEG has low compression efficiency and MPEG has high delay. While being a new gerneration video coding standard, H.264 is more suitable for the application to TV guidance for its high efficienct coding with low delay. In this thesis, the key techniques of H.264 video compression are studied to implement an image processing system for TV guidance. After the technique frameworks of many image and video coding standards are summarized, the comparison of them is also conducted. And especially the framework and principle of H.264 are analysed, focusing on the key module of intra-/inter- prediction, DCT transform and quantization, and entropy coding. It establishes a solid theoretical foundation for hardware design and realization.Texas Instruments’s TMS320DM642 DSP is specially designed for multimedia signal processing. DM642 with high CPU frequency has rich resources on the chip. The main research of the thesis is to design and realize a real-time dual-core video compression platform based on DM642 processor. On this board DSP and FPGA cooperate to capture video sequence, compress it using H.264 algorithm, and then transmit the code stream. In the receiver, DSP decompresses and displays the video sequence on the monitor. In the meantime,the circuitry contains Ethernet, VGA and PC104 interface, it can work on target tracking, pattern recognition and classification, so the system has a feature of versatility and extensibility. On the basis of the normal operation of the system, the thesis also does an optimization on system-level, function-level and sentence-level respectively. Finally the system can compress CIF image at the rate of 20 frames per second that meets the requirement of real-time video compression. The system can achieve PSNR of 33.72dB and good restoration images, at the compression rate of more than 20.Fast intra- and inter- frame prediction algorithm for image compression system is one of the key techniques of H.264. The thesis has conducted an in-depth analysis on existed algorithm theory for intra- and inter- frame prediction. To meet the needs of the fast prediction, the thesis does plenty of research based on the search algorithm. Proved by the simulation on PC programs, the modified algorithm improves much in the image compression efficiency and quality of the restoration results.