立体匹配中边界信息的强化算法  

The enhancement algorithm of the boundary information in stereo matching

作　　者：徐姗姗[1] 刘应安[1] 徐昇[1] 

机构地区：[1]南京林业大学信息科学与技术学院,江苏南京210037

出　　处：《山东大学学报（工学版）》2012年第6期43-49,共7页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金资助项目(30671639);江苏省自然科学基金资助项目(BK2009393)

摘　　要：为解决立体匹配中边界模糊、无纹理区域的边界误匹配、初始匹配代价过于粗糙等问题,设计了一种基于改进数据项的图像边界信息强化算法。通过归一化匹配代价削弱匹配不利的像素点;通过设置权值强化边界信息,对边界进行匹配约束。实验证明,新的能量函数能在边界区域得到正确视差;消除了大量无纹理区域的不连续线段;以较少的迭代次数提高了优化效果。在图片无纹理区域可降低14%的错误率。To solve the problem of fuzzy boundaries in stereo matching, boundary mismatching in textureless area and too rough cost in initial matching, a strengthen algorithm based on the boundary information was designed to improve the image of the data items. The unsuitable matched pixels were weakened by normalized matching cost, the weights were settled to strengthen the border information, and the boundary constraints were used to do the match. Experiments showed that, the new energy function could get the correct disparity in the boundary region and could eliminate the dis- continuity lines in the textureless region; the optimization effect was improved with less iterations. In the textureless are- a the error rate could be reduced to 14%.

关 键 词：立体匹配 能量函数 数据项 边界信息 边界约束 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]