Implementation And Optimization Of Windows Mobile-based Video Decoder

Modern digital video surveillance systems are mostly network based application systems that combine several technology fields including video processing, network communication, storage and display technologies. At present it is experiencing a transforming stage from analogous video surveillance to digital ones.Of all the video surveillance system present, it is most common that the surveillance client is based on non-portable devices such as PCs or TVs. Thus, it cannot fully take all the advantages of digital video surveillance. This makes a restricting factor the confines its deployment. This is also an outstanding advantage that smart devices have because they are easy to be carried around. But due to the highly computational complexity of the H.264 decoder, it is difficult to realize real-time video decoding on smart devices which have less computing and lower storage abilities. So it is a key step to optimize the decoder.This paper introduces a full set of decoder system that can run stably on windows mobile 5 platforms. The decoder is simply implemented and has simply interfaces for each module, which makes it maintainable. Synchronization and exception mechanism guarantees the stoutness of the whole system. Based on the detailed analysis of the decoder afterwards, besides the optimization functions the compiler provides, optimization methods are illustrated from system project level, C language level to assembly language level. First, based on the C's optimization restricts, from the perspective of the program's logic structure and C's characteristics, several optimization methods are given which fully take the advantages of C. Second, as to the key modules that have the most impact on the performance of the decoder, WMMX methods are interpreted in detailed, the use of WMMX fast improves the parallel of data processing. The adjustment of instruction order fully makes use of processor's pipelines mechanism, and increase the pipeline's throughput. At last a contrast result is given.It is demonstrated that the optimized decoder can run fluently on Windows Mobile, it can satisfy the basic requirements of mobile video surveillance.