Image fusion method based on self-learning neural unit

Abstract

The invention discloses an image fusion method based on a self-learning neural unit, and belongs to the field of image fusion. The method comprises the following implementation steps: 1) enabling a fused image to enter Mask R-CNN network for processing to obtain a corresponding mask image, a mask matrix, category information and score information; 2) constructing a self-coding network, and performing image feature selection, fusion and reconstruction by using a convolutional neural network (CNN); 3) carrying out sparse assignment on the convolution weight of the fusion layer, and adding minimum / maximum norm weight constraint and an L1 regular term; 4) respectively calculating the overall structure similarity SSIM, the regional structure similarity SSIM and the mutual information MI of thefused image and the source image; 5) training the neural network and adjusting parameters. According to the image fusion method based on the neural network, horizontal measurement and weight distribution can be obtained through learning network parameters in an optimal mode in a combined mode, the image definition is enhanced, the visual effect is improved, and the quality of a fused image is improved.

Description

technical field [0001] The invention belongs to the field of image fusion, and relates to an image fusion method based on self-learning neural unit, which is widely used in military, remote sensing, and computer fields. Background technique [0002] With the rapid development of image fusion technology and its wide application in military, remote sensing and other markets, multi-source image fusion technology has attracted the attention of researchers. In the past few years, various image fusion methods have been proposed. The traditional MST fusion method based on multi-scale transformation processes images in a multi-scale manner in the transform domain, specifically including solving the transform domain corresponding to the source image, fusing images according to artificially designed fusion rules, and inverse transforming to obtain the final image. Fusion methods include fusion methods based on Laplace Transform LAP (Laplace Transform), fusion algorithms based on non-...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art, and propose a self-learning neural unit for image fusion to solve the problems of activity level measurement and fusion rule design, and drive the fusion neural unit to combine in an optimal way through the loss function Obtain activity level measurements and weight assignments to enhance image clarity, improve visuals, and improve the quality of fused images

Images