Design And Realization Of Video Coding In High Definition Ip Camera Based On Stream Processor

With the development of the computer, network, communication and video compression technology, video surveillance technology has come to the digitalized and network developing period. Research of high-definition network camera has become the most popular technical studies; one of its key technologies is how to achieve the high efficient video signal processing in the system, such as video compression and transmission. How to encode high-definition video data in real-time would put heavy demands on processors. Stream architecture addresses large-scale parallel computing array and software managed memory hierarchy. Through exploiting on-chip parallelism and using application specific instruction, stream architecture provided area- and energy- efficient high performance in terms of architecture. The professor W.J.Dally of Stanford University introduced the world's first commercially available stream processor-Storm which has achieved good performance in the media and signal processing domains.This paper introduces Storm stream processor and AVS standard, and then explains our implementation of AVS encoder on Storm using stream processing and stream programming model based on stream architecture. Finally, this paper gives performance analysis of our AVS encoder. The main work of this paper is present as follow:Firstly, a novel rate control method for AVS is proposed, which is adapted to the video sequences with scenes change frequently. Experimental results show that the proposed algorithm can implement a smooth quantitative video quality and improve percept quality dramatically, especially have a better effect than JVT-G012 for the video sequenceds with scenes change frequently.Secondly, the stream processor architecture and stream programming model have studied in detail. This paper proposed AVS high-definition encoding structural design based on processor architecture and implemented AVS encoding algorithm on Storm stream processor using streaming programming mode. Experimental results show that our AVS video encoder can achieve up to 34fps for high-definition (HD) 720P (1280X720) image format which achieves high-definition real-time encoding.Finally, the programming optimization techniques of Storm stream processor and the evaluating limiting factors in application performance have been analysed in detail. For the stream programming, the organization of streams and the division of Kernel is the key and focus, which directly affects the other optimized performance. Double buffer technology can effectively hide the DMA latency of data to improve the task-level parallelism, greatly enhancing its overall performance. In view of powerful processing capability of DPU, the most important thing of design and optimization is to improve the efficiency of DPU as much as possible.