Stream processing model is a newly emerging parallel processing model, which is derived from media processing. However, as an important application of media processing, H.264 coding encounters problems during its streaming procedure. The main reason is that dependency limits the data level parallel exploitations. To solve this problem, data level parallel methods for different coding modules were developed by the analysis of coding algorithm. By adopting the method, key coding modules of H.264 can attain large data parallel granularity. Experiment results show that the data level parallel streaming inter/intra prediction, CAVLC, and the deblock filter can achieve outstanding acceleration.