一种鲁棒型Hausdorff距离图像匹配方法

An Image Matching Algorithm Using Robust Hausdorff Distance

图像匹配是图像处理的一项关键技术,传统方法受光照、噪声和遮挡的影响,使匹配过程变得困难。为了提高图像匹配的鲁棒性,提出了一种基于方向信息的鲁棒型Hausdorff距离匹配方法。该方法采用方向信息提取图像边缘,通过计算边缘匹配率(edge matching rate,EMR)获得候选匹配区域,然后采用修正后的Hausdorff距离构造相似性测度。实验结果表明,该方法加快了匹配过程,提高了抗噪性能,并能够准确匹配含有遮挡和伪边缘点的图像,从而解决了基于传统Hausdorff距离匹配方法因噪声点、伪边缘点和出格点而造成的误匹配问题。

Image matching is very important in image processing. The conventional matching methods are easily affected by occlusions ,light and noises ,which make the matching more complicated. In order to improve the robust performance of image matching ,this paper proposes an orientation-based robust Hausdorff distance for image matching. An edge detector based on the direction information is performed to get edge image. To achieve the promptness of image matching,EMR （edge matching rate） is introduced to determine which position can be the possible matching location, and further to construct a similarity measure based on the improved Hausdorff distance to find the correct location. The experimental results show that the proposed algo- rithm speeds up the matching process and it improves the resistance to noise. In addition ,this method matches the image occlusions correctly and overcomes the mismatching problems that induced by noise, spurious edge segments and outlier points, which demonstrate that the proposed method is feasible and effective.