适用于DIBR的匹配误差校正算法  

Correction Algorithm of Matching Error for DIBR

作　　者：张玲[1] 邰国钦[1] 刘然[2] 谢辉[1] 许小艳[2] 

机构地区：[1]重庆大学通信工程学院,重庆400044 [2]重庆大学计算机学院,重庆400044

出　　处：《华南理工大学学报（自然科学版）》2011年第12期51-55,共5页Journal of South China University of Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61071190);"211工程"三期创新人才培养计划建设项目(S-09102);重庆大学中央高校基本科研业务费专项资金资助项目(CDJZR10180013)

摘　　要：由于可见性变化、深度图像的不精确性以及计算的不精确等原因,采用基于深度图像绘制(DIBR)技术生成的虚拟视点视图中可能会存在匹配误差.为此,提出了一种适用于DIBR的匹配误差校正算法——零交叉校正(ZCC)算法.该算法建立在顺序匹配约束的基础上,通过分析目标图像像素点与参考图像上对应匹配点的映射关系来确定匹配误差区域,将区域内所有点的匹配点都赋值成起始点的匹配点,从而达到匹配误差校正的目的.实验表明,该算法可以有效地消除匹配误差.由于该算法与DIBR中的三维图像变换过程没有直接联系,因而它可以作为一种附加的误差检测与校正算法应用于DIBR的后处理过程中.Due to the variability in virtual visibility,the imprecise depth image and the inaccurate calculation,matching errors may occur in the virtual views synthesized by DIBR(Depth Image-Based Rendering) techniques.In order to solve this problem,a matching error correction algorithm named Zero-Cross Correction(ZCC) is proposed based on the order matching constraint.ZCC determines the matching error region by analyzing the mapping relations between the pixel points in the destination image and the corresponding matching points in the reference image,and it assigns the matching points of all points in the matching error region to the matching points of the starting point in the same region,thus successfully implementing the correction of matching error.Experimental results indicate that the proposed algorithm is effective in eliminating matching errors and that it can be applied as an additional error detection and correction algorithm to the postprocessing of novel views generated by DIBR because it is totally independent of the 3D image wraping in DIBR.

关 键 词：立体电视 基于深度图像的绘制 匹配误差校正 三维图像变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]