纹理图的3D-HEVC深度图编码单元快速划分算法

A fast coding unit size decision algorithm for the depth map based on texture in 3D-HEVC

为了降低3D-HEVC编码标准中深度图的帧内预测编码复杂度和编码时间,使用概率统计分析纹理图的最大编码单元(LCU)划分分布与深度映射图的LCU划分分布之间的相关性,提出一种快速算法,通过判断纹理图中LCU的划分深度,跳过深度映射图中一些不必要的LCU划分和模式判别过程,从而在基本不影响视频质量的前提下,提高编码效率。实验结果表明,可以在保证峰值信噪比(PSNR)和比特率(bit-rate)基本不变的情况下,大幅度减少深度图的编码时间。

In order to reduce the intra coding complexity and encoding time of the depth map in 3D-HEVC,a fast coding unit size decision algorithm is proposed based on the relevance between the largest coding units（ LCU） division depth and the LCU division distribution of the depth map in the probability statistics analysis texture map. This algorithm is to skip some unnecessary LCU division and pattern discrimination in the depth map according to analysis of the LCU depth in the texture map. In this way,the coding efficiency can be improved without affecting quality of video. The experimental results reveal that the proposed algorithm can save encoding time significantly on the basis of guaranteeing peak signal to noise ratio（ PSNR） and bit rate.