Research On Algorithms For Infrared And Visible Image Fusion Based On NSST

In recent years,with the development of electronic information technology,the way of acquiring signal has gradually developed from single sensor to multi-source sensor.In order to obtain more comprehensive scene information and make up for the lack of single sensor information,the features of each image of multi-source sensor has been extracted and fused.As an important branch of multi-source image fusion,infrared and visible image fusion has great application value in military,security monitoring,industrial manufacturing and other fields.The infrared sensor can detect the thermal radiation information of the object and capture the hidden target in the scene,but its image resolution is usually low;the visible light sensor imaging according to the reflected light of the object,has high image resolution and detail texture,but it is easy to be affected by the light conditions.According to the complementary characteristics between infrared and visible images,the fusion of the image can better retain the scene target information,which is conducive to the following image processing operations such as image detection and recognition.In recent years,the image fusion method based on multi-scale domain has been widely concerned by researchers at home and abroad because of its good ability of image decomposition and expression,which can better preserve the details of source scene contour and other information.In this thesis,the fusion algorithm of infrared and visible image based on multi-scale transformation is studied.Aiming at the problem that the traditional fusion algorithm can not keep the details of the target well and the image noise is easy to be introduced,the corresponding improvement scheme is proposed.The main work is as follows: 1.The image decomposition method based on multi-scale transform is studied.The image multi-scale decomposition algorithms such as Laplace transform,wavelet transform,double tree complex wavelet transform,NSCT(Non-Subsampled Contourlet Transform)and NSST(Non-Subsampled Shearlet Transform)are studied.The contrast experiments are carried out.2.Aiming at the problem that the traditional fusion algorithm is easy to lose the detailed information of scene object contour structure,an image fusion algorithm based on NSST and adaptive dual channel PCNN is proposed.In this algorithm,the infrared and visible images are decomposed by NSST,the low-frequency part of the image is weighted by gradient energy,and the high-frequency part is fused by dual channel PCNN with adaptive parameters,and the iterative threshold of the model is optimized.In order to further preserve the edge texture information of the fused image,an improved Laplacian energy sum is introduced into the high frequency fusion algorithm to enrich the edge texture information of the fused image.3.Aiming at the problem that original image noise is easy to be introduced in the fusion process,an image fusion algorithm based on NSST and guided filtering is proposed.In this algorithm,the infrared and visible images are decomposed by NSST,the low frequency image is fused by fuzzy logic,and the high frequency image is fused by a gradient domain guided filter with multi visual feature decision.In order to further improve the visual effect of the fusion image,the saliency image is introduced into the low-frequency fusion algorithm,and the saliency region is weighted fusion by statistical gradient energy of the saliency region.In this thesis,the fusion algorithm of multi-scale low-frequency and high-frequency parts is studied.Aiming at the problem that the traditional algorithm is easy to lose target information and introduce noise,the corresponding solutions are proposed,which lays the foundation for the follow-up research.