In image registration, the ratio of Euclidean distance from the closest neighbor to the second closest neighbor is used in SURF algorithm to find the matching point. When the matching threshold is very big, the method brings lots of mismatching points, which affect the accuracy of image mosaic. RANSAC is an effective method, which can eliminate the mismatching points, but it needs to make sure some parameters artificially. A method of multi-level image registration is put forward in this paper. It can not only eliminate the mismatching points, but also complete the image registration without participate of manual work. The experimental results prove the method of image registration can extract the matching point exactly, realize the image registration, and satisfy the need of image mosaic. It has a particular application value.%在图像匹配中,SURF算法采用最邻近与次邻近欧式距离比查寻匹配点,当阈值过大时,这种方法会产生大量误匹配点,从而严重影响图像配准的精度.RANSAC算法是一种有效剔除误匹配点的方法,但需要人工确定部分参数值.提出一种多层次图像匹配方法,不仅能够剔除误匹配点还可以无需人工参与完成图像匹配.实验表明,该匹配方法能够精确提取匹配点和实现自动匹配,很好地满足图像配准的要求,具有一定的应用价值.
展开▼