节点文献
基于DM642的MPEG4算法的移植和优化
Transplant and Optimization of MPEG4 Algorithm Based on DM642
【作者】 邵应昭；
【导师】 初秀琴；
【作者基本信息】 西安电子科技大学 ， 电路与系统， 2009， 硕士
【摘要】 随着多媒体信息需求的增长,在低速信道上实现实时动态视频编解码技术是目前视频处理领域的一个研究热点,嵌入式视频编解码技术有着越来越广泛的应用前景。本文实现了基于DM642的MPEG4嵌入式视频压缩系统,可应用于需要数据压缩的嵌入式多媒体领域。本文首先深入研究了DM642的硬件特性,包括处理器结构,视频口外设、存储器结构、并行硬件结构、集成开发环境CCS和DSP/BIOS的配置和使用方法,并分析了优化时可以利用到的所有硬件特性。针对视频压缩,重点研究了MPEG4算法的框架和流程、算法的各个组成模块、视频口的结构和视频采集功能的实现原理,为算法的移植奠定基础;其次,作为本文的重点,深入分析了算法移植中遇到的问题,并针对具体问题设计了优化中采用的各种方法,分模块逐一进行优化、针对每一种优化方法,测试了优化前后的性能并进行对比,同时列出了按一定流程进行优化后的性能测试总结。本文采用调整程序结构、存储器优化、流水线技术、线性汇编、内联函数、自带函数库等优化方法对移植成功的MPEG4 C语言程序进行优化。经测试,本系统优化前为2秒一帧,经优化后达到了CIF格式16帧每秒的压缩任务,基本满足了视频监控系统的需求。
【Abstract】 With the increasing demand on Multimedia information, It is a hot research area where implement real-time and live video coding on low speed transmission channel. Main harvest of this paper was that it introduced a MPEG4 embedded video compression system which would be normally used in embedded multimedia area who request video compression.First of all, this paper introduced hardware characteristic of DM642 which included the structure of the CPU and memory, video peripheric, the parallel hardware structure, and configuration and application of CCS and DSP/BIOS, and made a analysis of the hardware traits which was favor to optimization. Aiming at video compression, highlight of this part was focused on the process of implementing video capture and the principle and utility of video driver, this all made a basis for next job. In the second part of this paper, it covered the frame and flow of MPEG4 algorithm, and specially enhanced the modules which composed the algorithm. The last part, as the center of this paper, gave a detailed description of the obstacles encountered when transplanted and the methods used when optimization, and gave the comparison of every mothods that was been employed between the front and after condition of the optimization. Meanwhile, it demonstrated a performance-tesing summary that was arranged by the flow of optimization.After some optimization methods such as adjusting program structure and memory, pipeline technology, linear assembly, inline function and Image Library. After tset, it reflected a fact that the system completed a CIF format compression task of 15 frames per second compared with the previous 2 second per frame and basiclly satisfied video monitor system.