一种HEVC编码单元快速划分决策算法

A Fast Coding Unit Divided Decision Algorithm for HEVC

HEVC是下一代视频编码标准,为高清和超高清视频提供了编解码方案.在视频质量相同的情况下,HEVC比H.264节省约50%的有效码流.HEVC高的压缩率是以复杂的递归划分算法,更长编码时间为代价的.本文提出了一种基于统计的快速划分决策算法,旨在减少不必要的子划分,降低编码复杂度.该算法是利用CU空间相关性的特点:CU越小,划分后四个子CU就越相似,它们的率失真值就越接近CU率失真值的1/4.通过对CU的率失真值和划分前CU的率失真值的1/4比较,决定是否继续进行子划分.实验结果表明,该算法在码率增加不超过2%的前提下,在低复杂度配置和高效率配置下平均节省了24%和29%的编码时间,PSNR值基本不变.

HEVC is the next generation video coding standard, which provides the solution for high or ultra-high videos codec. It can save about 50% bitstream than H. 264 in the same quality of videos. HEVC high compression ratio is at the expense of more complex recursive partition algorithm and longer encoding time. This paper proposes a fast divided algorithm based on the Statistics that is aim at reducing unnecessary sub-division and cutting down the encoding complexity. The algorithm relays on the spatial correlation characters of the CUs. The four sub-CUs of the smaller CU are more similar and the rate distortion value of the sub-CU is closer to the quarter of CU rate distortion value. The algorithm concludes the division of the sub-CU by comparing the sub-CU rate distortion value with the quarter of CU. The experimental results show that our proposed method achieves about 24% time saving in low complexity and 29% in high efficiency configuration compared with the reference software on the condition that the bitrates increase are not more than 2% and PSNR change can be negligible.