Research On Image Denoising Algorithm Based On Non-subsampled Contourlet Transform And Statistical Modeling

Digital images in the transmission or conversion are polluted by noise unavoidably,which leads to the distortion of the image information.So image denoising is of great significance in the image processing.In order to get better results in the research of image processing,such as feature extraction and pattern classification,it is imperative to eliminate the noise in the noisy image.The non-subsampled Contourlet transform(NSCT)has many properties of Contourlet transform.Meanwhile it also has complete shiftinvariance.There is no pseudo Gibbs effect in image reconstruction.Thus,NSCT is widely applied to image processing.In this paper,on the basis of non-subsampled Contourlet transform,two new algorithms for image denoising are proposed.The work of this paper is summarized as follows:(1)Image denoising by employing various threshold functions can cause pseudo Gibbs effect and inherent deviation.In order to overcome these shortcomings,the new image denoising algorithm based on non-subsampled Contourlet transform and improved adaptive threshold are proposed in this paper.Firstly,the non-subsampled Contourlet transform is applied to the original image.Secondly,the new threshold function and the adaptive threshold with considering the energy distribution are obtained aiming at the shortcomings of the traditional threshold function and universal threshold.Thirdly,the NSCT coefficients are rewritten by using the new threshold function and adaptive threshold.Finally,the image is reconstructed by the inverse non-subsampled Contourlet transform.The experimental results show that the proposed algorithm obtains better visual effects and denoising results compared with the state-of-art denoising algorithms.(2)The researchs show that the coefficients of the non-subsampled Contourlet transform have a certain correlation.According to the distribution characteristics of NSCT coefficients,a statistical model is established accurately.An image denosing algorithm based on Multivariate Generalized Gaussian distribution(MGGD)and adaptive wiener filter in NSCT domain is proposed in this paper.Firstly,the vectors of the image can be estimated by MGGD model that can be adjusted by the parameters.Meanwhile,the parameters of MGGD and the NSCT coefficients of the original image can be obtained by the adaptive adjustment of the minimized residual and the maximum a posteriori(MAP).Finally,coefficients after denoising are filtered by the adaptive wiener filter.The experimental results show that the algorithm obtains higher Peak Signal to Noise Ratio(PSNR)compared with the state-of-art de-noising algorithms,and keeps the edge of the image better.