A Self-Supervised Learning Fusion Method for Multiband Images

Abstract

The invention relates to a multi-band image fusion method, in particular to a multi-band image self-supervised fusion method based on a multi-discriminator generative confrontation network. The method is carried out according to the following steps: designing and constructing a generative confrontation network. It is composed of a generator, and the label image is the multi-band source image itself; the generator network structure is composed of two parts, a feature enhancement module and a feature fusion module, and a generative model is obtained through the dynamic balance training of the generator and the discriminator, and the multi-band image is obtained. Fusion results. The invention realizes the neural network of end-to-end self-supervision fusion of multi-band images, and the result has better clarity, amount of information, richer detailed information, and is more in line with the visual characteristics of human eyes.

Description

technical field [0001] The invention relates to an image fusion method, in particular to a multi-band image fusion method, in particular to a multi-band image self-supervised learning fusion method. Background technique [0002] At present, wide-spectrum multi-band imaging has been widely used in high-precision detection systems, and the existing research is mainly carried out on the infrared and visible light bands. Therefore, it is imminent to explore the simultaneous fusion of multiple (≥3) images. In recent years, the research on image fusion based on deep artificial neural network has emerged. However, due to the lack of standard fusion results in the field of image fusion, that is, the use of deep learning to establish image fusion models generally lacks labeled data, resulting in difficulties in deep learning training or poor fusion effect. The larger the number of synchronously fused images, the more prominent the problem is. [0003] Self-supervised learning is one...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem of limited fusion results due to the lack of label images when using deep learning methods in the field of image fusion to fuse multi-band images, the present invention proposes a new method for self-supervised learning and fusion of multi-band images based on multi-discriminator generation confrontation network

Images