Three-dimensional deformed face limitation-based face pore matching method

Abstract

The invention discloses a three-dimensional deformed face limitation-based face pore matching method. The method comprises the following steps that: two face images of the same person under differentangles are inputted, and each face image is preprocessed; a pore level scale invariant feature transform transformation algorithm is adopted to extract the features of the pore scales of the obtainedface images, so that the skin pore features of a plurality of pore feature points in the face images can be determined; five-sense organ positioning and three-dimensional deformed face fitting are performed on the obtained face images, three-dimensional deformed faces are embedded into the two face images through fitting; three-dimensional deformed face limitation is performed on the fitted face images; the key points of the extracted features of the pore scales are matched with the key points of the pore scales of the limited three-dimensional deformed faces; and a final matching result is obtained. With the method of the invention adopted, other screening steps are not required, matching is very precise, the accuracy of the matching can be improved, and the establishment of an accurate face pore database can be benefitted.

Description

[0001] technology neighborhood [0002] The invention belongs to the field of computer vision research, and relates to image processing, three-dimensional face technology and deep learning method, in particular to a face pore matching method based on three-dimensional deformed face restriction. Background technique [0003] The face pore feature, like the face pupil feature and fingerprint feature, is an effective identity authentication feature. In the pore feature matching neighborhood, there are skin pore feature matching based on the whole face, and skin pore feature matching based on local areas, but the above matching methods all have certain shortcomings, the matching area is too large, and the pores are in the image. The colors in the image are similar, resulting in insufficient matching accuracy; the matching area is too small, and the number of corresponding feature points is small, resulting in insufficient matching accuracy. [0004] If the general full-image feat...

AI Technical Summary

Problems solved by technology

But the human face is non-rigid, and it cannot be calculated simply by mechanical methods.

Images