Deeply Pipelined DSP Solution to Key Modules in H. 264

With the rapid development of microprocessor, embedded multimedia products are gradually becoming the mainstream in the market. However, the high coding efficiency enabled by the H.264 video compression standard comes with substantially greater algorithmic complexity as compared to that of existing standards. And this additional complexity results in many difficulties in the implementation and optimization tasks. This paper analyzes the algorithms of the two time-consuming modules of integer transform and motion estimation in H.264. Then optimizes the two modules based on the extended instruction set of C64x/C64x+. Finally, deeply pipelined DSP solutions to two modules are presented in this paper. The experiment results show that optimizing parallel assembly can make the codes more efficient.