MPEG-4 Algorithm Transplantation And Optimization On DM642 Platform

With the development of mobile Internet and muti-media technology, the embedded video encoding equipments have been widely used. MEPG-4 as a very mature video coding standard, which has a moderate algorithm complexity, good compression performance and excellent portability in the embedded platform, has been widely used in network video and other fields. The latest embedded video encoding equipments are required to have not only video compression capabilities, but also video analysis functions, which require an embedded platform encoder has lower computational complexity, and can provide some interfaces for video analysis. To achieve these demands in this paper, algorithm transplantation and optimization for MPEG-4 are implemented. Finally, real-time video encoding and transmission are completed on DM642.Firstly, this paper analyzes the principles of MPEG-4 technology and studies the key characteristics of DM642 platform. Transplantation is done and practical demand is accommodated.Secondly, in the algorithm, this paper mainly optimizes the motion search module which has the highest computational complexity. An improved algorithm is proposed, which is based on the existing fast motion estimation algorithm. The algorithm can greatly reduce the number of motion search by using multiple motion vector prediction, early termination and new search template.Thirdly, in the optimization level, the framework of encoder has been adjusted according to the DM642 features. The core modules of the encoder, such as SAD and quantification, have been rewritten with linear assembly language and TI library, which greatly enhances the encoder’s efficiency.The paper also designed the camera driver, which achieves the camera data collection and YUV422 to YUV420 video format conversion.To achieve network communication, the video data packets of UDP encapsulation is redefined. Only minimal CPU resources are required when transmitting video data stream.The final experimental result shows that the optimized MPEG-4 algorithm can achieve 4CIF real-time video encoding on DM642 platform and transmit compressed video data with network at the same time.