基于多视点视频整帧丢失的自适应错误掩盖算法

发布时间：2018-02-26 01:11

本文关键词： 错误掩盖多视点视频压缩(MVC) 外推块组合　出处：《光电子.激光》2014年03期 　论文类型：期刊论文

【摘要】：为解决多视点视频压缩(MVC)导致视频整帧丢失风险增大的问题,基于MVC的特点提出了一种4方向和多参考帧的错误掩盖算法。首先,在同一视点内,将相邻帧沿其各自运动矢量外推,得到视点内的最佳运动矢量;在视点间采用同样的方法,得到最佳视差;以外推的重叠面积为判断依据,在视点内错误掩盖和视点间错误掩盖之间选择较优的掩盖方式;最后,在MVC已有块模式的基础上引入新的块组合模式,合并具有相同错误掩盖方式、运动矢量的绝对差值小于阈值的相邻块,以减小图像的块效应。实验结果表明,与已有算法相比,本文算法有效利用了MVC的内部机制,改善了重建帧的客观指标,增强了主观质量。
[Abstract]:In order to solve the problem that multi-view video compression can increase the risk of whole frame loss, a 4-direction and multi-reference frame error concealment algorithm is proposed based on the characteristics of MVC. By extrapolating adjacent frames along their respective motion vectors, the optimal motion vectors in the viewpoint can be obtained; the optimal parallax is obtained by using the same method among the viewpoints; the overlap area of extrapolation is the basis of judgment. Finally, a new block combination mode is introduced on the basis of the existing MVC block pattern, and the merge has the same error concealment mode. The absolute difference of motion vector is smaller than the adjacent block of threshold value to reduce the block effect of image. The experimental results show that compared with the existing algorithms, the proposed algorithm makes effective use of the internal mechanism of MVC and improves the objective index of the reconstructed frame. Enhanced subjective quality.
【作者单位】：天津大学电子信息工程学院;
【基金】：国家高技术研究发展计划(863计划)重大项目(2012AA03A301) 国家自然科学基金重点资助(60932007,61201179) 教育部博士点基金(20110032110029) 天津市科技支撑计划(11ZCKFGX02000)资助项目
【分类号】：TN919.81

【参考文献】