基于3D-HEVC标准的相邻块视差矢量获取算法质量优化的研究

Quality Improved Neighbor Block Disparity Vector Derivation in 3D-HEVC

在高效视频编码标准的三维扩展方案(3D-HEVC)中,相邻块视差矢量获取算法是视点间纹理图编码的一项关键技术.针对现有快速算法获取的视差矢量质量有所下降的问题,本文提出一种利用分组搜索和按比例组合策略的相邻块视差矢量获取质量优化算法.首先,改变原有相邻块视差矢量获取算法中将第一个搜索到的视差矢量作为最终视差矢量的准则;其次,设计一种新的分组结合方法,将时间方向和空间方向上的待搜索的候选块分为若干小组,并将互相临近的候选块中得到的视差矢量按比例结合为新的视差矢量;最后,设计新的分组搜索策略,在新的视差矢量与原有矢量中获取最终相邻块视差矢量.实验结果表明,在维持现有基于编码单元的简化相邻块视差矢量获取算法效率的前提下可实现编码质量提升.

Neighboring block disparity vector derivation in 3D High Efficiency Video Coding（ 3D-HEVC） is a key technology to get the disparity vector. To improve the quality of the disparity vector derived from the fast algorithm,we propose a proportion combination and group searching method to get the NBDV. We first changed the rule that takes the first search DV as final NBDV,and then divide the searching candidates into several groups for the proportion combination,after that design a strategy to search the candidates in several groups. The simulation results showthat the proposed algorithm has improved the coding quality with negligible coding time loss.