基于分割的立体视频运动估计方法  

Video segmentation based motion estimation for stereoscopic video

作　　者：南立园 张萍[1] 王璟璟[1] 夏清[1] 费春[2] 

机构地区：[1]电子科技大学光电信息学院,成都610054 [2]电子科技大学计算机科学与工程学院,成都611731

出　　处：《中国科技论文》2017年第2期174-179,共6页China Sciencepaper

基　　金：国家自然科学基金资助项目(61308102);高等学校博士学科点专项科研基金资助项目(20130185120038);四川省科技计划基金资助项目(2016GZ0089;2014GZ0005);中国博士后科学基金资助项目(2013M531946)

摘　　要：随着新型采集、显示器件的出现,带有场景信息的多视点彩色帧加深度信息的立体视频成为研究热点。随之而来的是更大的数据量,立体视频运动估计是去除视频信息冗余的重要方法。传统的运动估计方法将图像分为固定大小的子块,通过求取参考帧和当前帧对应子块的最小绝对误差(sum of absolute difference,SAD)来确定该块的最佳运动矢量。由于存在子块的错误匹配现象,传统的运动估计方法得到的内插帧有明显的模糊和块效应。为了克服固定子块引起的缺陷,将视频分割应用到运动估计中,提出了1种基于视频分割和深度约束的立体视频运动估计方法。实验结果表明,用分割一致性来限制块匹配过程,可以有效提高运动估计精度,从而使内插帧在主客观都得到改善。With the emergence of new stereo video sensor,display devices,stereo video with multi-view color and depth frames contain the scene information becomes a research hotspot.Motion estimation for stereo video is a major method to solve the issue of large data through transmission.In traditional methods,frames are divided into fixed sized blocks,then the best motion vector of the block is defined by calculating the minimum（sum of absolute difference,SAD）corresponding to the sub block of reference frame and the current frame.The interpolated frame using traditional approach usually presents obviously blurring and blocking artifacts because of the sub-block matching error.In order to overcome the problem,a novel motion estimation algorithm based on video segmentation and depth constraint for stereo video is proposed.Experimental results show that the proposed algorithm can significantly improve both objective and subjective the accuracy of block matching and the motion vector field,and provide the better performance in interpolated frames.

关 键 词：图像通信 立体视频 运动估计 视频分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]