Study Of Optimized H.264 Decoder On Embedded Multiprocessor

As the original video data is so voluminous for real applications, video coding has been a focus of research and applications for saving transport bandwidth and storage space. For now, many video coding standards have been established for different applications, such as MPEG-1/2/4, H.261/H.263. In the latest, the just released H.264 and AVS standard have achieved significant improvement compared with previous video coding standards, but they have similar framework! So it is very important for us to have an in-depth study on the key techniques used in video coding, its realization under different hardware environment and performance optimization according to different environmental characteristics.In this paper, realization of H.264 decoder on the embedded multi-processor, OMAP5912, is the focal point. According to H.264 profile and level suggestion and handset characteristics, H.264 Baseline profile is chosen. The application focus of multi-processor lies in the conversation mechanism between its kernels, which is how the kernels carry out the different tasks in the high efficient way. Based on the introduce of RF6, which is the OMAP5912 multi-kernel conversation mechanism, this paper also recommend the parts that compose H.264 decoder, and single out the parts which are fit for the single kernel, such as the file operation management, parsing micro-block mode, creating the image pixels data by predict data and residual data, are suit for ARM processor; CAVLC decoder, IDCT, inverse quantization, intra prediction, inter prediction, deblocking filter, are suit for DSP processor.The paper mostly narrates theory and realization of the parts transplanted to DSP, consisting of CAVLC decoder theory, sequence of coefficients decoder; theory and realization of Hadamard and integer IDCT, inverse quantization; intra prediction mode and inter motion estimation mode, pixel interpolation for different pixel precision, MV prediction mode; boundary strength, adaptive filter, general and special filter mode for deblocking filter; frame structure of whole system. Finally based on the results of the test sequence such as foreman, bus, football, coastground on OMAP5912, the conclusion that OMAP5912 can achieve 34fps for QCIF, 11fps for CIF is educed.In conclusion, summery is derived, which also involves weak points of this paper, and look forward to the future insight for the embedded multi-processor in the wireless communication and multimedia processing.