An infrared and visible light image fusion method based on special information synchronous decomposition

Abstract

The invention provides an infrared and visible light image fusion method based on special information synchronous decomposition. The infrared and visible light image fusion method comprises the following steps of image preprocessing, image mean value stripping, dictionary learning and construction, image joint sparse decomposition, sparse coefficient fusion, mean value fusion and image reconstruction. According to the invention, an image fusion algorithm for carrying out synchronous orthogonal matching pursuit decomposition on specific information of infrared and visible light images is provided under a joint sparse representation model, and the original fusion mode according to the image block can be further refined to be fused with all atom representation coefficients participating in the image block. The fusion method has the advantages that the fusion comprehensiveness and integrity are improved, the obtained fusion image is obvious in characteristic and comprehensive in information, that is, the specific information of the infrared image and the visible image can be fused together, the fusion image is clearer integrally, and a certain area distortion phenomenon cannot occur.

Description

technical field [0001] The invention relates to the field of image fusion, in particular to an infrared and visible light image fusion method based on synchronous decomposition of unique information. Background technique [0002] Due to the difference in imaging principles between infrared sensors and visible light sensors, the image information acquired by the two sensors is complementary and redundant. Infrared imaging is capable of detecting concealed or deliberately camouflaged targets, and works 24 hours without interruption, but its spatial resolution is poor; visible light images are imaged according to the reflection information of light, and have a certain spatial resolution. Most of the information in the scene is Can be clearly seen, its detailed information is richer [0003] In recent years, the existing image sparse representation model method has received extensive attention. This model represents the image signal by the linear accumulation of a small number ...

AI Technical Summary

Problems solved by technology

Infrared imaging is capable of detecting concealed or deliberately camouflaged targets, and works 24 hours without interruption, but its spatial resolution is poor; visible light images are imaged according to the reflection information of light, and have a certain spatial resolution. Most of the information in the scene is Can be clearly seen, its detailed information is richer

[0003] In recent years, the existing image sparse representation model method has received extensive attention. This model represents the image signal by the linear accumulation of a small number of atoms in the redundant dictionary. The non-zero coefficients of these atoms represent the main feature structure of the image. It is widely used at present. In the field of image fusion, although this method achieves synchronous decomposition of multi-source images, it does not strip the redundant information of multi-source images, which is not conducive to image fusion. Therefore, on the basis of sparse representation, the joint sparse representation model is proposed The feature extraction problem of multi-source images is solved. The model defines redundant information and complementary information of multi-source images as common information and unique information respectively. Although the joint sparse representation can better extract the feature information of infrared and visible light images, but The decomposition process is still solved according to the OMP algorithm, and the fusion rule is still the processing of the unique sparse coefficient vector. For the unique information to be fused, the two may not use different atoms in the dictionary when performing sparse decomposition, which is similar to the The source image is decomposed using different wavelet bases, so it is essentially two different decomposition methods, which makes it difficult to perform coefficient trade-off and fusion, and cannot achieve simultaneous decomposition

Images