Abstract: Because of the lack of texture region is short of distinctive features or having serious repetitive features, the matching algorithm purely based on feature is very difficult to obtain the strong unique feature descriptors, which leads to greatly increasing of mismatching ratio. But the AKAZE algorithm based on the nonlinear scale space, which can better take into account the texture and detail information of image, so it has a certain application value for image matching in the lack of texture region. Calculating the cross-power spectrum between images by Fourier-Mellin phase correlation technique using the FFT and log-par conversion, can obtain the rotation, scale and translation parameters between the two images. So, this paper obtains the geometric relationship between the two images by Fourier-Mellin conversion, which is used for restraining the matching range of the feature points detected by the AKAZE algorithm, to solve the problem of image matching in lack of texture region. Experimental results show that this method can increase the number of matching points, and the matching accuracy.

Key words: Fourier-Mellin conversion, AKAZE algorithm, the lack of texture region