Research On Visible Light Image Restoration Technology Based On Infrared Features

Images,the most intuitive data type since the information age,are also one of the important ways of expression for people who communicate.However,in the process of image acquisition,problems with low-light scenes inevitably occur.The images collected in this scene are characterized by low brightness,color deviation,low contrast,and obscure details,which not only reduce people's vision,but also affect the post-processing of the image.Therefore,it is necessary to study the restoration technology of images in low-illuminance scenes in order to post-process the images and help people extract important information.Aiming at the visible light image acquired under low illumination,the thesis used the infrared image characteristics in the same scene and combines the fusion-based method to process the low-illuminance image,so that the image information and colors in this scene are richer,and the purpose of restoring the image is achieved.As a result,an improved Retinex enhancement algorithm was used to enhance and preprocess the low-illumination visible light image,and then image restoration processing was performed based on an algorithm that combines HSV space with dual-tree complex wavelet transforms.The specific research contents are as follows:1)From the perspective of Retinex enhancement algorithm,a Retinex enhancement algorithm based on adaptive guided filtering was proposed.On the one hand,guided filtering was based on the principle of locally variable linearity and had edge retention characteristics and gradient retention characteristics,but there were still some shortcomings.To solve this problem,adaptive guided filtering was introduced to estimate the brightness component,which not only made the details of the image more obvious,but also avoided the halo artifacts of the color image.On the other hand,when the traditional Retinex algorithm was based on the RGB channel to enhance the image,the color image would be distorted.In order to solve this problem,the strategy of optimizing the color reduction factor and normalization function to adjust the brightness was adopted,which not only solved the problem of color images that were prone to distortion,but also solved the problem of low image brightness.The simulation results show that the algorithm in the thesis is compared with some existing algorithms,the enhancement effect is more obvious,and it has better practicability.2)From the perspective of image fusion algorithm,an image fusion algorithm based on HSV space and dual-tree complex wavelet was proposed.First of all,because the color information and brightness components of the HSV space were separated from each other and were suitable for image processing,the thesis processed the image based on the HSV space.Then,the dual-tree complex wavelet transform not only had the advantages of the wavelet transform,but also overcome the shortcomings of the wavelet transform's directional limitation and poor translation.Therefore,the dual-tree complex wavelet was used to decompose the two images in the frequency domain.Finally,in the fusion rule,a fusion strategy of adaptive principal component analysis weighting was used to fuse low-frequency sub-images,and a fusion strategy that combines local energy and matching degree was used to fuse high-frequency sub-images.Through simulation experiments on three sets of images,the results show that compared with other algorithms,the algorithm in the thesis has moderate brightness,richer details,better results,and better robustness.