Infrared visible light image fusion method based on 3D convolutional neural network

Through the infrared visible light image fusion method based on 3D convolutional neural network, the problem that image fusion strategy in the prior art is difficult to adapt, and efficient information retention and enrichment of image fusion results are achieved.

CN119722489AActive Publication Date: 2025-03-28WUHAN POLYTECHNIC UNIVERSITY
5 Cites 0 Cited by

Patent Information

Application Number
CN202411886963.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-20
Publication Date
2025-03-28
Estimated Expiration
2044-12-20

AI Technical Summary

Technical Problem

In the prior art, when infrared visible light images are fusion, it is difficult to adaptively adjust the fusion strategy according to the input image, resulting in neutralization or supersaturation of image brightness, loss of information, and affecting the fusion performance.

Method used

The infrared visible light image fusion method based on 3D convolutional neural network is adopted, and the high-dimensional semantic features are extracted through the encoder. The feature fusion module performs cross-match stitching and weighting processing, and the decoder performs weighting fusion to obtain the image fusion result.

Benefits of technology

The correlation between feature maps is enhanced, the information content of the feature map is enriched, the loss of effective feature information is avoided, and significant thermal targets and rich texture structures can be adaptively retained, and the fusion results of retaining more effective information in the source image are obtained.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to an infrared visible light image fusion method based on a 3D convolutional neural network, and belongs to the technical field of image enhancement, and the method comprises the steps: obtaining a training complete image fusion model which comprises an encoder, a feature fusion module and a decoder; obtaining an infrared image and a visible light image to be fused; performing high-dimensional semantic feature extraction on the to-be-fused infrared image and the to-be-fused visible light image based on an encoder to obtain a plurality of visible light feature maps and a plurality of infrared feature maps; performing cross matching and splicing on the plurality of visible light feature maps and the plurality of infrared feature maps based on a feature fusion module, and weighting spliced features to obtain a processed feature map; and carrying out weighted fusion on the processed feature map based on a decoder to obtain an image fusion result. According to the invention, the technical problem that the fusion performance is affected due to brightness neutralization or supersaturation, information loss and the like in image fusion in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art