| With the development of society,the state has paid more and more attention to the military and security fields.The fusion of infrared and visible light images as a favorable tool in this field has also received more and more attention.The pixel-level image fusion technology mainly has two key links.One is the construction of the transform domain.The image is transformed from the spatial domain to the frequency domain.The image is often more detailed and specific in the frequency domain than in the spatial domain.Therefore,the image is in the frequency domain.The fusion effect is better.The second is the selection of fusion methods.According to the characteristics of the images,the appropriate fusion rules are selected to achieve a better fusion effect.At present,most of infrared and visible light fusion images have low contrast and unclear scene detail information.The algorithm of this thesis deeply explores two key links of image fusion and proposes two new fusion algorithms.The main results are as follows:1.For the existing infrared and visible fusion images,there are low contrast,untargeted target information,and unclear background details.This paper proposes a new algorithm for infrared and visible light image fusion in the NSCT domain.In this method,the source image is first decomposed by using NSCT at different scales and in different directions to obtain low-frequency subbands and high-frequency subbands.The low-frequency subbands are fused using sparse representation-based fusion rules,and the high-frequency coefficients are improved by using improved pulses.Coupled neural network(PCNN)fusion rule fusion.In this paper,we use the improved Laplace and Log-Gabor energy to replace the traditional fixed value as the PCNN link strength,and the average of the two link strength output results is taken as the final PCNN output,and finally passed through the inverse.NSCT obtains fused images with high contrast and rich details.2.This paper introduces a multi-wavelet-V system with detailed mathematical expressions.Using the multi-resolution of V-system and the multi-directionality of NSCT,the infrared images and visible light images are decomposed in multiple levels and decomposed to obtain low frequency coefficients.Fusion rule fusion based on sparse representation,high-frequency coefficients are merged with fusion rules based on two-dimensional Log-Gabor energy,detail information is merged with improved pulse-coupled neural network fusion rules;finally,the fusion image is obtained through corresponding inverse transformation.The algorithm decomposes the source image from different levels and directions,so that the details of the source image are meticulously characterized.Combined with multiple fusion schemes,the detailed information of the fused image is more clear and the objective indicators are also significantly improved. |
