Research On Image Denoising Methods Based On Contourlet Transform And Shearlet Transform

The image is the main source of human to get information. However, the imageacquisition, transmission or storage process is often polluted by noise, which hasserious implications for the subsequent image processing. Therefore, image denoisingplays a vital role in image preprocessing. Wavelet transform is widely used in imageprocessing because of its good time-frequency and multi-resolution characteristics.However, along with the deepening of research, Wavelet transform gradually exposedits flaws and shortcomings. Wavelet transform can only reflect the point singularityand have not the anisotropy. Multiscale geometric analysis method is good to make upfor the inadequacy of wavelet transform. In this paper, the main object of study is amulti-scale geometric analysis. Image denoising algorithm based on Contourlettransform and Shearlet transform was proposed. The main contents and research workis as follows:(1) The background and significance of the image denoising was described, andwavelet transform and multi-scale geometric analysis are briefly introduced. Then theadvantages and disadvantages of each method were analyzed and summarized. Thesubsequent and new image denoising algorithm was proposed based on thetheoretical basis(2)A new threshold was proposed by analyzing the flaws of the general threshold,and an improved threshold function was proposed to make up for the inadequacy ofthe hard threshold function, the soft threshold function, the half-threshold functionand the semi-soft threshold function. We proposed an new image denosing algorithmwhich was combined by the new threshold and improved threshold function based onthe improved Contourlet transform. The feasibility and effectiveness of the new imagedenosing algorithm is verified through numerous experiments.(3)Considering the inadequate performance of Contourlet transform sparserepresentation, Shearlet transformation was studied. We combine normal inversegaussian model Because it can accurately modeling of arbitrary degree of tailingsignal.An image denosing algorithm was proposed based on the Shearlet transformand normal inverse Gaussian model. By comparison with several types of classicaldenoising algorithms, peak signal-to-noise ratio (PSNR), mean structure similarity(MSSIM) and visual aspects are greatly improved.