Efficient realization of parallel HEVC intra encoding

The emerging next generation video coding standard - HEVC (High Efficiency Video Coding) achieves more than 50% bitrate reduction in comparison with its predecessor H.264/AVC under the same visual quality, but at the expense of a substantial complexity increase. As a result, a traditional single thread encoder which operates in sequential way is difficult to be optimized to a satisfactory level, for example, a real-time HEVC codec for Ultra High Definition resolution (4Kx2K@50fps and beyond). Thus parallelism in the encoding stage must be extensively utilized to meet such requirements. In this paper, we give a full analysis of fine granularity parallelism for HEVC intra coding and exploit its maximum parallel degree in context of coding unit dependency. Then an efficient parallelization encoding strategy is proposed and implemented on x265, the first open source HEVC encoder targeted to real-time application, to demonstrate its potentiality. Experimental results show that our scheme could achieve average 502% speedup gains for different video sequences without performance loss.