高清视频并行处理的研究

A study on parallel processing for high-definition video

本文基于一款具有并行运算能力的数字信号处理器,完成了一个高清视频处理系统.该系统中,高清摄像采集芯片输出的视频图像采用Bayer格式,数字信号处理器获取视频后,通过已申请专利的基于绝对时刻块截断编码变换的Bayer图像压缩算法进行压缩.此算法综合考虑了运算复杂性和压缩率,使得在现有平台和网络上实现高清视频传输成为可能.该系统中,高清摄像采集芯片可直接与数字信号处理器的视频端口连接,无需外部解码器,这样既提高了系统的集成度,也降低了开发的时间和成本.为了提高代码的稳定性和可维护性,数字信号处理器程序采用了一种专为视频处理算法设计的程序框架,所以不必考虑底层驱动和任务间协作的可靠性问题.该系统可用于交通自动化等领域,具有低成本、可靠性高等特点.

High-definition(HD) video processing is always in company with high bandwidth and enormous calculated amount.In order to reduce bandwidth,new video compression algorithms with high compression ratio continue to emerge.However,improvement in compression ratio causes increasing complexity of algorithm.Therefore,how to improve computational efficiency has become a hot research topic.To achieve this goal,there are many methods,among which parallel computing is a very important one.It has lower cost to improve system performance.Because of the importance of parallel computing,many processors have their own structures to achieve parallelism,and parallel processing capacity has become an important measurement of the performance of a processor.In this study,we make full use of a digital signal processor(DSP) with parallel computing capabilities to construct a high-definition video processing system.We accomplished both hardware and software development of the system.In the hardware structure,a high definition video capture integrated circuit(IC) is used,which is able to output HD Bayer format images directly.The DSP we use could deal with Bayer format itself,therefore the capture IC could connect to the DSP video port directly,without the need of additional transformation.This hardware structure does not only improve the system's integration,but also reduce the development time and costs.We developed a new algorithm which is based on absolute moment block truncation coding to compress Bayer images.The algorithm makes fully consideration of the complexity and the compression ratio.It is a balanced solution.It is able to reduce the signal bandwidth,so that the output video stream is suitable for network transmission.What's more,this algorithm is computational affordable.The Bayer compression algorithm can reduce the bandwidth which is essential for network transmission.In order to improve stability and maintainability of the software program,the program uses a framework which is specially developed for video processing program.Thus,we don't need to consider the underlying drivers and the reliability of collaborations between tasks.This system can be used for traffic surveillance system and other fields,and it can both reduce the cost of monitoring systems and improve reliability.