基于投票加权累积度量的模板匹配算法

从点集相关性的角度提出了一种新的模板边缘图像匹配度量——投票加权累积度量(WVAM),在该度量中融入了抗几何畸变以及抗杂点与相似区域干扰的机制,能够实现异源情况下模板边缘图像的匹配定位。为了进一步提高WVAM匹配的单相关峰特性,转换点的坐标投票为局部结构信息投票,形成了融入局部结构相似性的投票加权累积度量(LSS-WVAM),该度量能够表征模板边缘图像与待匹配区域的整体结构相似性,更具有稳健性。在仿真实验中利用全局与局部度量信噪比作为评价指标,证明了WVAM具有比LTS-HD(Least trimmed square Hausdorff distance)更好的全局单峰与局部梯度特性。与WVAM相比,LSS-WVAM在全局和局部性能上约提高30%和4%。

Template matching based on weighted voting accumulation measure

A novel edge template matching measure is proposed based on the point set correlation, termed weighted voting accumulation measure (WVAM). The measure is capable of resisting the interference of noise and the similarity region. In order to further improve single correlating peak of the measure, the voting is based on the local structure information instead of the point coordinates, and thus a weighted voting accumulation measure based on the local structure similarity (LSS-WVAM) is proposed. LSS-WVAM represents the structure similarity between the edge template and match region, and thus is robust. In experiments, global and local signal to noise ratio is used as the evaluating criterion. And the experimental results illustrate WVAM has better characteristic in terms of global single peak and local gradient than LTS-HD(Least trimmed square Hausdorff distance). Furthermore, LSS-WVAM can improve the global performance 30% and the local performance 4%.