Image registration method and system based on Gaussian field constraint and manifold regularization

Abstract

The invention provides an image registration method and system based on Gaussian field constraint and manifold regularization. The method is used for establishing correct matching by removing the wrong matching in the initial matching point pair, includes establishing a non-rigid geometric transformation model corresponding to the geometric transformation between images to be matched, establishingan objective function based on Gaussian field constraint and manifold regularization, and solving a parameter matrix by a deterministic annealing method; calculating the non-rigid geometrical transformation model by using the model parameters, and judging the correctness of the initial matching point pair according to the threshold value; and finally, establishing the mapping relationship betweenthe two images by using the correct set of matching points, and realizing the image registration. The invention models the non-rigid transformation between the images to be registered, and establishes a robust estimator, which greatly reduces the error rate of the matching, and still maintains good robustness even in the case of a large number of error matches in the preliminary matching.

Description

technical field [0001] The invention relates to the technical field of image registration, in particular, the invention relates to an image registration technical solution based on Gaussian field constraints and manifold regularization. Background technique [0002] The basic goal of image registration is to correspond to the same parts of the body. By finding a spatial transformation, the points corresponding to the same position in space in two images of the same scene obtained by different sensors at different times or angles of view are aligned. One-to-one correspondence, so as to achieve the purpose of information fusion. [0003] In the past decades, scholars have studied many methods to solve the image registration problem. These methods can be roughly divided into two categories: region-based registration methods and feature-based registration methods. The former searches for matching information by searching the similarity of the original gray value in a certain a...

AI Technical Summary

Problems solved by technology

[0007] Although these methods have been successful in many fields, when the image contains a lot of local distortion and the image content is complex, many wrong initial matching point pairs will be obtained after the initial matching. When the error rate exceeds a certain percentage, these methods cannot Effectively remove errors

Images