Research On The Method Based On The Curvelet Image Denoising

For the electronic file of "scratch" and "stain" noise in AIC Image Scanning System, two new improved image denoising methods in a Curvelet transform domain is presented.There are a circulation translation image denoising method based on the statistical feature of isotropic and a circulation translation image denoising method based on the statistical feature of anisotropy. Through the analysis of the statistical distribution characteristics of Curvelet transform coefficients, with isotropic model and the model for all the members of the opposite sex to approximate statistics. Under the bayesian maximum a posteriori criterion expression of coefficient estimate is derived. At the same time, model of complex has led to cannot obtain the approximate solution of the estimated coefficients. Using the Newton--Raphson numerical method,the estimated parameter approximation is got. In addition, in order to effectively eliminate noise due to the discontinuous points to the pseudo Gibbs phenomenon, is also used in algorithm of the circular translation links, further improve the quality of the restored image. The experimental results show that denoising methods in this paper the peak signal-to-noise ratio(PSNR), respectively, compared with the traditional Curvelet denoising method on average 2.5 d B and 3.1 d B. And they can avoid reconstruction images appear in the "scratch" and "embedded stain", while effective denoising, can better protect the image edges and details...