Research On Image Denoising Model Based On Fractional Calculus

The main reason for the noise in the image is the interference of the random signal when the image is captured or transmitted,which is an influence factor that hinders people from collecting the correct information of the image.Noise is everywhere in real life.People have to deal with the noise in the image in order to obtain a clear and effective image,which gives birth to a series of digital image processing technology.Although it has been developed for many years,the exploration of a new generation of image denoising methods still does not stop,and it is still an important topic in the field of digital image processing.With the emergence of the definition of fractional calculus,it is found that the application of partial differential equation and convolution mask model to image denoising can well eliminate the influence of noise,which promotes the research and development of partial differential equation and mask model.this paper is mainly based on fractional calculus image denoising model research,the main work and achievements are as follows:1.Aiming at Gaussian noise,a fractional partial differential equation image denoising model with fractional fidelity term is proposed.After adding fractional fidelity term,this model can eliminate the ladder effect produced by other models such as Perona-Malik(PM)model and Rudin-Osher-Fatemi(ROF)model,because the existence of this fidelity term makes the whole image appear smoother,and more detail textures in the original image can be saved.By comparing with other algorithms,the effect of image denoising is measured by evaluation indexes such as peak signal-to-noise ratio and structural similarity,and the optimal value is selected iteratively,so as to retain the denoised image with the best effect.2.Aiming at the speckle noise in medical images,a convolution mask image denoising model based on adaptive fractional calculus is proposed.Self-adaptation in the model algorithm is mainly reflected in the construction of an exponential parameter relationship which is closely related to the image,which can adjust the parameter values dynamically,which makes the model algorithm more practical.The model mainly uses four steps to remove speckle noise.First,select the appropriate membership function model and construct the weight table of the image with the help of the membership function;second,construct the convolution template of 5x5 according to the definition of calculus,and construct the exponential adaptive parameter function relation by analyzing the relationship between the convolution template parameters and the gradient information of the image;third,select the window of 3x3 and use the weight table to traverse the whole noise image.Finally,the corresponding convolution template generated by the function relationship is convoluted with the traversed noise image to get the final denoised image.Compared with the known algorithm model,this model has a good effect on the removal of speckle noise,and can better preserve the texture information of the image to a certain extent.