Image one-way point-to-point matching method based on minimum deformation energy

Abstract

The invention relates to an image uni-directional pointwise matching method based on minimum deformation energy. First, the feature point of each profile of an image to be matched is marked in sequence; second, between a template image and the image to be matched, the one having more profile feature points is chosen to be a reference image and the one with less profile points is set to be a imageto be aligned, and mutually corresponding start points are set in the two images; third, the minimum deformation energy is taken as a basis for determination to find all profile feature points corresponding to all profile feature points of the reference image in the image to be aligned, wherein, if the (n)th profile feature point in the reference image corresponds to the (n-1)th in the image to be aligned, then when the deformation energy of the (n+1)th profile feature point in the reference image is calculated, only the (n-1)th and the (n)th profile feature points in the image to be aligned are taken into consideration; and forth, the sum of the minimum deformation energy is calculated according to the deformation energy of the mutually corresponding profile feature points and a judgmentthat whether the sum exceeds a matched threshold value is made so as to determine whether the two images match with each other. The method has less amount of calculation, consumes less resource and is capable of validating the minimum deformation energy.

Description

Technical field [0001] The invention relates to the application field of graphics and image matching, in particular to a one-way point-by-point image matching method based on minimum deformation energy. Background technique [0002] Graphic image matching has important theoretical significance and application value in computer applications such as three-dimensional reconstruction, target recognition, content-based image retrieval, text recognition, and building recognition. And shape matching is one of the most widely used and studied matching methods. The current matching methods are mainly divided into two categories: one is shape matching based on invariants, which includes methods based on global geometric features, such as the invariance of geometric features such as area, circumference, compactness, eccentricity, etc., to match images ; Also includes methods based on transform domain features, mainly various invariant moments, Fourier operators, wavelet descriptors, morpho...

AI Technical Summary

Problems solved by technology

[0003] At present, the more commonly used matching method is the vertex correspondence method, that is, inserting a vertex on the contour line of the image, that is, breaking the contour line, so that the vertex in one image corresponds to a point on the contour line in another image, but In this way, nonlinear interpolation will evolve, the calculation cost is very high, and it is difficult to verify that the final result is the minimum deformation energy

Images