Replantation And Refining The Predictive Coding Mode Of Digital Video Processing Based On Embedded System

Digital video processing plays an important role in the fields of industry, military enterprises, civil facilities and so on. However, the old standards of video processing can not meet the rising market demand which results in the emerging of a new generation standard represented by MPEG-4. New application is brought by the new standard. The sources of signal are becoming more and more complicate and application platform is expanding. Under the circumstance, the researching nowadays focuses on how to apply the new application on various platforms and how to enhance the application quality. XVID is a set of open-sources tools used for digital video processing based on MPEG-4. It provides main coding tools as below: interframe coding, intraframe coding, predicting direct current coefficient and alternating current coefficients, nonrestraint motion vector, motion estimation, predictive motion vector field adaptive search technique, and four motion vectors. This set of coding tools, characteristic of high efficiency and replantablity, is especially suitable in video for network and video for wireless. Accordingly, the author takes XVID as the researching example.Predictive coding model is one of the core models of MPEG-4. Its performance directly affects the whole capability of the processing procedure. Predictive coding model consists of motion estimation and motion compensation. The function of predictive coding model is to predict the present processing frame through the processed frame, thus aiming at showing image frame of digital video by using data as few as possible. During the digital video processing, most image frames apply the mode of predictive coding in coding and even the image frames that do not use the predictive coding still apply the mode of predictive coding in coding direct current coefficient and alternating current coefficients after their discrete cosine transformation. Predictive coding model is the foundation and key of the digital video processing. The author will take predictive coding model as the core of the researching.The research on application of embedded system is a hotspot of researching nowadays. Many present advanced applications of digital video take embedded system as the platform. To promote innovation and practicability of the task, the author will take embedded system as the ultimate realization platform and choose the representative digital signal processor TMS320C6416 produced by America Texas Instruments Ltd. Co. as the example.First, the author analyzes the kernel processes of XVID according to the theory of digital video processing and MPEG-4. Second, the author discusses about replanting this set of tools to TMS320C6416 platform after a good command of XVID. Third, aiming at predictive coding model of XVID, the author devises a scheme of refining including two levels: using theory of variable complexity algorithm in refining based on algorithm irrespective to platform; using characteristics of realization platform in refining based on instruction pertinent to platform. A series of algorithms devised in refining based on algorithm are based on practice observation and theory exploration, which completely cultivates the inherent characteristics of digital video and cooperates with present fast algorithm. The application areas are expanded and the processing rate is raised. Meanwhile, the image quality is not damaged. A series of codes of functions devised in refining based on instruction are based on characteristics of TMS320C6416 and use developing and optimizing tools offered by environment of compiler. Processing rate is promoted and image quality is not damaged. This set of optimization scheme is not only the researching tool of the author but also contributes a lot to the application of digital video on other platforms.