Research On Image Fusion And Multiplicative Denoising Method Based On Total Variation

Image denoising is a basic step in image processing.In reality,except the standard Gaussian additive noise,multiplicative noise is also widespread.For example,in the field of synthetic aperture radar imaging,the coherent combination of several radar echoes in each resolution unit results in the serious speckle phenomenon.In SAR images,the noise obeys the gamma distribution of the unit mean,and the reduction of speckle is a long-term theme.In addition,due to the diversity and complexity of image data sources in recent years,image fusion technology has attracted much attention.The goal of image fusion is to extract and synthesize significant information in the original images,usually significant features are considered as the basic structure,edges,textures,etc.of the image.First of all,we find that multiplicative noise has a great influence on the high grayscale region of the image.Then we propose a grayscale detection operator and construct an ?-total variation regularization term,combined with the fidelity term suitable for gamma multiplicative noise,a convex adaptive total variational multiplicative denoising model is obtained.The existence and uniqueness of the model solution and the comparison principle are proved.Finally,using the framework of Chambolle's original-dual algorithm to solve numerically,and comparing it with the classical denoising model,the effectiveness of this model is shown.In the field of image fusion,the fusion of SAR images and optical images has been a research hotspot.In this papper,we study the fusion of images containing multiplicative gamma noise and images containing additive Gaussian noise,a new image fusion and denoising model is proposed.Then,two numerical solution methods are shown.One is a variational PDE method and is solved numerically using the finite difference method.The other is augmented Lagrangian method based on the concept of operator splitting,and the influence of model parameters on the fusion effect is discussed.For multi-sensor image fusion,compared with the visualized enhanced fusion model,this model has certain advantages in both fusion effect and computational efficiency.