Study Of Image Fusion For Visible And Near Infrared

Digital cameras that capture color images in low-light conditions need to set high gains or lengthen the exposure time.Such settings will produce noise and motion blur if the generated noise and blur are large,the effect of using the single image denoising and image restoration is not satisfactory.In recent years,studies on the use of nearinfrared to improve the quality of color images in low-light conditions have been carried out internationally.The basic idea is to extract brightness information using near-infrared images with little or no noise and to use noise-containing color images to remove noise.Extraction of color information,through the fusion of brightness and color information,retains the detailed information in the near infrared image while preserving the color information in the visible light image.The main work of this thesis is as follows:(1)Studying the BM3 D algorithm to denoise the image in low-light conditions.Using the image of the image database,the Gaussian noise of different intensities is added to simulate the images collected in the low illumination environment,and then conduct an experimental study on the most popular BM3 D denoising algorithm in the world.Experimental results show that the denoising effect of BM3 D algorithm is superior to other denoising methods,but with the continuous increase of noise intensity,the denoising effect of BM3 D decreases significantly.We improved the method to solve this problem.(2)Studying a fusion algorithm called Make My Day for near-infrared and visible light images in the low-light conditions.In order to solve the problem of poor singleimage denoising,we propose a near-infrared and visible-light image fusion algorithm based on Make My Day to enhance image texture details.Using a laboratory-built,simultaneous noise-free color image and near-infrared image experimental platform,we used the proposed algorithm to experiment with noise-free color images acquired in low-light conditions.Experimental results show that the proposed algorithm is significantly better than the single image BM3 D algorithm.Especially in the high noise environment,our algorithm can still have very fine texture details,and with the increase of noise,the effectiveness of the proposed algorithm does not drop significantly.(3)Improving the Reinhard algorithm for near-infrared and visible light image fusion algorithms in low-light conditions.In the process of fusion between visible light image and near infrared image,color distortion is easy to occur.In order to solve this problem,a weighted transfer algorithm for Reinhard algorithm is proposed.The global information of the brightness of the near-infrared image and the visible light image is collected,and the detail information of the near-infrared image is retained while the color information of the visible light image is preserved as well.Experimental results show that the image after fusion has a higher degree of clarity,more colorful,more detailed information,and is more suitable for the computer to do a variety of processing.