Image registration method and system based on feature-oriented gmm and edge images

Abstract

An image registration method and system based on feature-oriented GMM and edge images, including calculating the edge images of the two images to be matched, extracting the feature point set of the edge images, and setting the membership degree of the GMM model based on the feature description vector set. The model corresponding to the geometric transformation between the images to be matched is used to solve the model parameters through the optimization method; the posterior probability of the matching point of the feature point in the feature point set and the feature point in the other feature point set is recalculated, and the threshold is used to determine whether the two match each other. point. The present invention models the situation where affine transformation exists between images to be matched. Compared with the previous optimal method, the method is more robust and accurate, and is especially stable when data degradation is severe.

Description

technical field [0001] The invention relates to the technical field of image feature matching, in particular, the invention relates to an image registration technical solution based on a feature-guided Gaussian mixture model (GMM) and an edge image. Background technique [0002] The basic goal of image matching is to match the same parts of two images of the same scene obtained by different sensors at different times and perspectives. In many matching scenarios, due to factors such as the image sensor and imaging technology used, the environmental conditions of image acquisition, and the change of the scene itself over time, the images used for matching often have problems such as serious degradation and changes in details. Consistent feature reduction is a challenging problem for the registration of such images. In this invention, we focus on the problem of registration of more degraded images, where two or more images of similar scenes with different states are required, ...

AI Technical Summary

Problems solved by technology

In many matching scenarios, due to factors such as the image sensor and imaging technology used, the environmental conditions of image acquisition, and the change of the scene itself over time, the images used for matching often have problems such as serious degradation and changes in details. Consistent feature reduction is a very challenging problem for the registration of such images

For images with serious image quality degradation, it is difficult to extract reliable regional features

Images