基于傅里叶梅林变换的SURF算法

针对加速鲁棒性特征算法,在没有后续去误匹配等处理的情况下，对存在较大旋转角度的两幅待匹配图像，匹配精度较低的问题，提出了一种基于傅里叶梅林变换的SURF算法。该算法通过对待匹配图像和参考图像实施傅里叶变换和梅林变换，利用能量谱求出两幅图像发生的旋转角度，并利用SURF算法找出图像间发生的平移和尺度变化，实现了图像间的匹配。实验结果表明，该算法可有效地实现图像间存在较大旋转角度时的几何匹配，且相比已有的SIFT和SURF算法，具有更好的匹配效果。

A SURF Algorithm Based on the Fourier Merlin Transform

A SURF algorithm based on the Fourier Merlin transform is proposed to improve the registration accuracy of speed up robust features in dealing with the registration when there is large rotation between images under the condition that no removal error registration algorithm. The floated image and reference image are Fourier transformed and Merlin transformed by the proposed algorithm, after which the rotation angle between two images is computed by using the energy spectrum, and the changes of translation and scale between the two images are computed by the SURF algorithm, thus realizing the registration of two images. Experimental results show that the registration with large rotation between two images can be effectively realized by the proposed algorithm with better registration result than that by the SIFT algorithm and the SURF algorithm available.