Based On Tms320dm642 Video Processing System Design And Algorithm Research

With the rapidly development of information technology, video processing technology has been widely applied in daily life, including industry, medical treatment, military and others. Each video processing platform has a lot of differences because of different application fields and functions, so founding a universal video processing platform and doing further development on it according to various requirements has become a significant way of video processing platform design. The TMS320DM642 Digital Media Processor is considered to work well for people who use DSP at present, which is a 32-bit fixed-point DSP that is developed for multimedia applications by TI. It achieves the demand of video processing with their good processing performance and abundant on-chip peripherals. So TMS3420DM642 has become one of the perfect platforms for the video processing system currently.There are three parts in this thesis including design of the hardware of digital video processing system based on TMS320DM642, the development of the digital video processing driver and the scheme of application program, and the research for digital video segmentation algorithms and the optimization ways for the algorithm realization.At first, the author designs, researches and analyzes the development and implementation methods of the embedded video process system. The kernel structure and peripheral equipment of TMS320DM642 was discussed. On the foundation of research and taking into account the design requirement of this system, the implementation method in which the main design of video system based on TMS320DM642 is discussed, and the detail design is given.On the basis of the hardware selection and the system design, the thesis analyzes the structure and functions of the kernel of DSP/BIOS, it also discusses the components and work principle of RF5 Reference Framework. Based on this Reference Framework, we develop the common video processing driver on DM642 and bring forward the method that constructs the application program.After the study of the hardware and software for the system, this thesis does deeply research in the digital image segmentation algorithms based on DSP embedded system and the author puts forward a new image segmentation algorithm based Minimum-Description-Length that can fit the real-time demand of DSP system very well. In addition, the thesis does a research on several optimization ways of DSP programming and memory optimization technology.At last, the thesis brings forward several constructive advices on using the system to develop the algorithms and improving the system applications.