抑制编码误差扩散的深度图帧内编码

Intra-Frame Coding of Depth Maps to Restrain Coding Error Pervasion

为进一步提高深度图编码的质量和三维视频虚拟的合成质量,提出了抑制编码误差扩散的深度图帧内编码方法.首先根据深度图的特性将深度图划分为平坦区域和边缘区域,并分析这两个区域的误差扩散特性.基于此特性,在宏块的率失真优化计算中,利用修正值修正易引起误差扩散的块或子块,使其后续大量受误差扩散影响的块或子块具有较低的率失真代价函数.实验结果表明,与JMVC8.5的帧内编码方法和基于分割的帧内预测方法相比,文中方法虚拟合成的平均峰值信噪比分别提升0.30和0.25dB.

In order to improve the coding quality of depth maps and the virtual synthesis quality of 3D videos,an intra-frame coding method to restrain the coding error pervasion is proposed.In this method,the depth map is firstly segmented into a flat region and an edge region according to the features of depth maps.Then,the distortion perva-sion properties in the two regions are analyzed,based on which the blocks or sub-blocks that cause the distortion pervasion are corrected by using a correction value during the rate-distortion optimization of macro blocks.Thus,a large number of following blocks or sub-blocks influenced by the distortion pervasion may possess a smaller value of rate-distortion cost function.Experimental results indicate that,as compared with the intra-coding method presented by JMVC 8.5 and the intra-frame predication method based on segmentation,the proposed coding method improves the average peak signal-to-noise ratio of the virtual synthesis view by 0.30 and 0.25 dB,respectively.