基于傅里叶-梅林变换的图像匹配方法研究

摘要

图像匹配是计算机视觉中的重要应用之一，相位相关算法是其中的一种常用方法。但是该算法只能准确匹配存在平移的情况，对带有旋转和缩放的情况无法正确匹配。针对这一问题，本文提出了基于傅里叶-梅林变换的匹配方法，利用图像傅里叶变换幅度谱中的包含的信息，计算出旋转角度和缩放系数，根据结果变换模板图像，进一步计算出平移量。本文设计了模拟和实际实验对算法进行验证。结果显示本文算法不仅在模拟条件下具有较好的匹配效果，在实际情况下也能准确匹配。%Image matching is one of the important applications in computer vision, and a commonly used algorithm of this is phase correlation algorithm. This algorithm can only accurately matches images when the shift is included,but it is not correct with the rotation and scaling. To solve this problem, this paper proposed a matching algorithm based on Fourier-Mellin transform. The spectrum of image’s Fourier transform contains information of the rotation angle and scaling factor. Algorithm in this paper can calculate these by this information. Template image is transformed by these results. Then, image shift is calculated by phase correlation algorithm. Simulation and actual experiments are designed to verify the effect of algorithm. Results show that our method not only has a good match under simulated conditions, but also works well in practice conditions.