A Study Of Image Denoising Algorithm Based On Finite Ridgelet Transform

Image denoising is an important part of the image preprocessing process.The effect of image denoising directly affects the efficiency and precision of the subsequent related work of image processing.In order to guarantee that the results of subsequent image processing are more reliable,it's required to retain more image edge and texture information when perform the image denoising.The finite ridgelet transform is used to image denoising for its high performance of maintaining the linear features of the image and enhancing the visual effect of the image.However,the finite ridgelet transform is achieved by the Radon transform which leads to “Wrap Around” phenomenon when image is reconstructed,and results in the edge blur after the image denoising,that is to say the image noises haven't been removed completly.In this paper,we propose two novel image denoising algorithms to address the problem of edge blurring in traditional finite ridgelet transform;The first is based on finite ridgelet transform and ROF model(FRIT_ROF).The second is based on the finite ridgelet transform image denoising method(m Res FRIT_ROF)of multi-level residual filtering.The details of the two algorithm are described as follows:(1)FRIT_ROF algorithm is an improvement to the traditional finite ridgelet denoising algorithm which includes two improvements.First,aiming at the shortcomings of the soft and hard threshold function in the denoising process,this paper proposes an adaptive bivariate threshold function.The threshold function value fluctuates between soft and hard thresholds,and its first and second order can be used to solve the edge oscillation.In this paper,the ridgelet coefficients are optimized in the process of threshold shrinkage due to the "Wrap Around" phenomenon caused by radon transform in ridgelet denoising.The minimum variation of the ridgelet coefficient is less than the threshold,rather than the traditional zero setting operation.Finally,the effectiveness of the FRIT_ROF algorithm is analyzed and demonstrated by designing multiple sets of experimental comparisons..(2)m Res FRIT_ROF is a denoising method based on multi-level residual image of the FRIT_ROF.The traditional ridgelet threshold denoising algorithm only considers the image information recovered in the noisy image.However,the residual image of the denoised image hasn't been taken into account.The residual image of the ridgelet threshold denoising still contains valuable image information.This paper proposes a new method of multi-level residual image denoising.The residual information of the image is reconstructed by BM3 D method and processed by Gaussian filters to serve as a compensation image fed back to the denoised image.Finally,the feasibility and effectiveness of the algorithm is demonstrated on multiple contrast experiments.