抑制编码误差扩散的深度图帧内编码

发布时间：2018-03-28 09:29

本文选题：编码误差　切入点：三维视频　出处：《华南理工大学学报(自然科学版)》2014年01期

【摘要】：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.
[Abstract]:In order to further improve the quality of depth map coding and the quality of 3D video virtual synthesis, a depth map intra coding method is proposed to suppress the spread of coding errors. Firstly, the depth map is divided into flat region and edge region according to the characteristics of depth map. The error diffusion characteristics of these two regions are analyzed. Based on this property, in the rate-distortion optimization calculation of macroblock, the block or sub-block which is easy to cause error diffusion is corrected by modifying the correction value. A large number of blocks or sub-blocks affected by error diffusion have lower rate-distortion cost functions. The experimental results show that compared with the intra coding method of JMVC8.5 and the intra prediction method based on segmentation, In this paper, the average peak signal-to-noise ratio (PSNR) of virtual synthesis is increased by 0.30 and 0.25 dB, respectively.
【作者单位】：西安电子科技大学综合业务网理论及关键技术国家重点实验室;
【基金】：国家自然科学基金资助项目(60902081,61001205) 高等学校学科创新引智计划项目(B08038) 西安电子科技大学中央高校基本科研业务费专项资金资助项目(72105457,72115612,K5051301020,K5051301027)
【分类号】：TN919.81

【参考文献】