采用光照不变特征改进支持权重快速立体匹配  

Fast stereo matching via illumination invariant features and modified support weights

作　　者：刘天亮[1,2] 戴修斌[1,3] 霍智勇[1,2] 朱秀昌[1,2] 刘峰[1,2] 

机构地区：[1]江苏省图像处理与图像通信重点实验室,南京210003 [2]南京邮电大学通信与信息工程学院,南京210003 [3]南京邮电大学地理与生物信息学院,南京210046

出　　处：《仪器仪表学报》2012年第9期2030-2036,共7页Chinese Journal of Scientific Instrument

基　　金：国家青年科学基金(610011523;1200747);国家自然科学基金(61071166;61071091;61172118);江苏省自然科学基金(BK2010523);南京邮电大学校科研基金(NY210053;NY210069;NY210073;NY211030);江苏高校优势学科建设工程资助项目

摘　　要：为了确定同一场景在不同视点下图像点的对应关系,提出一种基于光照不变特征和改进型支持权重核的快速立体匹配。首先,稠密构造光照不变局部特征计算初始匹配代价;基于三次立方Tri-Cube权重核双通聚合得到稳健的匹配代价;采用优胜者全选法选择初始视差。然后,利用基于多方向邻域Tri-Cube权重核的双边滤波、遮挡检测和误码掩盖等策略实现视差智能修复。实验表明,该方法能产生较为满意的稠密视差图,计算费用低,能提高对光照变化的适应能力。A fast stereo matching method based on illumination invariant features and modified support weights is proposed to determine the correspondences in different images of the same scene taken from different viewpoints. Firstly, illumination invariant features are constructed densely for the stereo pairs and initial matching costs are calculated from the invariant features; robust matching costs are obtained based on two-pass aggregation with tri-cube weight kernels; and initial disparities are selected using Winner-Takes-All optimization method. Secondly, intelligent disparity inpainting procedures are in turn implemented or executed with modified bilateral filtering based on multi-directional neighbor tri-cube weight kernels, occlusion detection and disparity error concealment. Experiments indicate that this technique with low computational burden, can produce comparably satisfactory dense disparity map, and can improve the adaptability to illumination variations.

关 键 词：立体匹配 光照不变量 特征描述符 三次立方权重核 多方向双边滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]