Abstract: For the occlusions and noise of contours matching，this paper introduces a new method which uses the mutual information as the two binary-images similar criterion and traverses all the candidate solutions to find the biggest mutual information and its location in the source image，this method has more matching accuracy than others.To accelerate the matching process，this paper introduces a new fast method for binary-image mutual information matching，the new method does not rely on the size of module and the source image，and only relies on the number of 1-valued pixels.Theory illustrates that the new method is really fast and experiment also shows that one matching only needs 50 ms，so the new method is effective.

摘要： 针对轮廓匹配中存在的遮挡、噪声等问题，提出把互信息作为二值图像相似性的度量准则，通过遍历的方法找到最大互信息的位置；为了加快匹配的速度，提出了相应的互信息计算的快速算法。新算法不依赖模板和输入图像的大小，仅仅与模板和输入图像中轮廓像素点的个数有关。实验结果表明，文中算法在512×512图像上完成一次匹配平均需要花费约50 ms，可以克服光照、遮挡、随机噪声和小形变的影响。