一种基于可操纵金字塔的鲁棒图像匹配算法

图像匹配是计算机视觉和图像处理领域中一个比较活跃的研究课题。针对离散小波变换对于图像平移和旋转的敏感性，提出了一种基于可操纵金字塔的图像匹配算法。算法考虑的是图像之间存在平移和旋转变换的情况。首先，变换的近似旋转角在图像分解的最底层采用环映射算法得到，然后在图像分解的其余层提取兴趣点作为匹配特征，并使用Hausdom距离来度量参考图像和待匹配图像之间的相似性，依次得到由粗到细的匹配结果。实验表明该算法具有较高的匹配精度和较强的鲁棒性。

A Robust Steerable Pyramid-Based Image Matching Algorithm

Image matching is an active topic in computer vision and image processing. The steerable filter was used to over-come the translation-and-rotation-sensitive of the wavelet transform. The rigid transformation was considered in the proposed algorithm. The rotated angle was computed approximately by ring-projection method at the lowest level of image' s decomposition. The interesting points were extracted as image feature and the Hausdorff distance was used to measure the similarity between the reference image and input image in the rest level of image's decomposition to refine matching result. Through a coarse-to-fine fast procedure, the final matching results could be achieved. Experimental results verify the algorithm is rather accurate and robust.