Image Registration Method Based on Fourier Mellin Transform and Maximum Mutual Information Theory

Abstract

The invention discloses a Fourier transform and powell algorithm based image registration method which mainly solves the problems that the Fourier-Merlin transform based registration accuracy is not high and the powell algorithm based registration is sensitive to an original value. The Fourier transform and powell algorithm based image registration method comprises (1) inputting a float image and a reference image; (2) performing low-pass filtering processing; (3) obtaining a transformation parameter through Fourier-Merlin transform; (4) calculating a mutual information value; (5) judging whether the mutual information value is the belt or not according to the mutual information and the powell; (6) continuing the search if not and repeating processes of space geometric transformation, statistics of area-color histogram, calculation of mutual information and optimized judgment; (7) searching a parameter satisfying the accuracy requirement; (8) performing geometric transformation; (9) performing image fusion. The Fourier transform and powell algorithm based image registration method overcomes the disadvantage that a Powell optimization algorithm is sensitive to the original value and obtains a good registration accuracy.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a high-precision registration method in the field of image registration, which first obtains initial parameters by rough matching. The method can be used for the registration of the medical image after affine transformation and the radar aperture synthetic SAR image. Background technique [0002] In recent years, image registration has been used in various target recognition and image fusion fields, and the application of evolutionary algorithms to image registration is a hot issue. From the perspective of the registration process, the most important part is to find accurate and appropriate registration parameters, and use the registration parameters to perform appropriate spatial transformation on the image, so as to obtain the correct registration result. [0003] At present, according to the registration process, it can be divided into feature-based image regis...

AI Technical Summary

Problems solved by technology

[0006] (PV) is obtained, and then Powell is used to search for registration parameters. Since the Powell algorithm has the defect of relying on the initial point, the search will fall into a local extremum, or the registration will fail.

Images