Multi-source image matching method combined with local phase sharpness directional description

Abstract

The invention provides a multi-source image matching method combined with local phase sharpness directional feature description, aiming at the matching problem caused by large intensity difference and nonlinear radiation distortion among multi-source images. Firstly, an image pyramid scale space is constructed, phase consistency solution is carried out on a frequency domain of an image on the basis, a maximum moment feature is obtained, and a KAZE operator is adopted to extract feature points. Then, Fourier transform is carried out by using a Log-Gabor even symmetric filter, improved local phase sharpness features and phase orientation features are constructed respectively, gradient amplitude and gradient direction features of the image are replaced in sequence, and a local phase sharpness orientation descriptor is established by combining a log-polar coordinate description template; and finally, similarity measurement is carried out by using the Euclidean distance to obtain corresponding points.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a multi-source image matching method combined with local phase sharpness directional description. Background technique [0002] As a window for human visual perception of the world, images are widely used in real life. With the rapid update of image sensor and photographic imaging technology and equipment, multi-source image data acquisition is becoming more and more abundant, and its processing has become a research hotspot. In order to meet the diverse requirements in the fields of object detection, scene recognition, and data fusion, it is first necessary to solve the matching problem of multi-source images. The essence of multi-source image matching is the process of obtaining corresponding points of images captured by different sensors. However, due to the different imaging mechanisms of sensors, multi-source image data has problems such as intensity differences, ...

AI Technical Summary

Problems solved by technology

The image matching method of deep learning is fast and has strong feature learning ability. However, due to the large differences in ground features between multi-source images and the difficulty in obtaining training samples, the generalization ability and applicability of this type of method are limited.

[0004] It can be seen that the intensity difference and nonlinear radiation difference of multi-source image matching lead to the problems of sparse corresponding points and low matching success rate still exist. Therefore, how to effectively overcome the existing problems of multi-source images, reduce image gradient sensitivity, and realize Image Robust Matching Has Practical Research Value

Images