Research On Image Denoising And Enhancenment Based On Curvature Filter

Image denoising and enhancement is an important branch in image processing.The main purpose of image denoising is to restore the original information of the image from the noise image,which is convenient for the subsequent work of image processing.Image enhancement selectively emphasizes or suppresses information in the image according to user needs,removes the blur of the image,highlights the local details of the image,and further improves the image quality.Using appropriate methods to remove noise and image enhancement is an important process for image processing.Therefore,effective removal of image noise while preserving the edges and details of images has become a hot topic at present.Gong et al.creatively proposed the curvature filtering method that implicitly calculates the curvature of the image and relaxing the smoothing constraints on the image,by constructing the projection operator.This image denoising algorithm is faster than the traditional variational denoising model,but there are also some obvious shortcomings.The focus of this paper is to study and improve the curvature filtering.The main research contents and innovations are as follows:(1)This paper starts with the image noise model and the classical image denoising method,introduces the basic principle of image denoising,elaborates the denoising idea of curvature filtering,and evaluates the peak signal-to-noise ratio and structural similarity.The indicators show the advantages of the curvature filtering model compared to the classical denoising method.(2)In this paper,the problem of incomplete image denoising by the curvature filtering model under strong noise conditions is studied.Based on the work(1),this paper improves the projection operator in the curvature filtering algorithm and enhances the denoising ability of the curvature filtering algorithm.Firstly,the implicit curvature calculation method using curvature filter replaces the calculation of higher-order partial differential equations in the traditional variational model,which solves the problem of complicated calculation and slow convergence to some extent.Secondly,when implicitly calculating curvature,use half window triangular tangent plane combination minimum triangulation plane projection operator replaces the minimum triangular tangent plane projection operator of curvature filtering to improve the denoising performance.The experimental results show that the proposed method can achieve a good balance between noise removal and detail information preservation under strong noise conditions.(3)In this paper,the problem of unable adaptive image denoising by the curvature filtering model under strong noise conditions is studied.On the basis of work(2),the adaptive strong noise denoising algorithm based on curvature filtering is proposed.Firstly,we considered the regular energy function of the curvature filter model,the prior estimation of the image is difficult to satisfy under the strong noise condition,when curvature is used as a regularization condition in the original curvature filtering.The local variance is used to distinguish the flat area,edge area and noise in the image.Then,according to the characteristics of strong noise spots in strong noise images,the regular energy function is modified to increase the regular energy of local variance,which makes the constraints of the regular terms more reasonable and improves the denoising performance of the algorithm.Experimental results show that PSNR and SSIM of the proposed adaptive denoising algorithm are higher than other comparison algorithms in this paper,which achieves better denoising ability and protects the edges and details of the image,and obtains better visual effects.