Image Denoising Algorithm Based On Contourlet Transform

Contourlet transform is not only good at properties multidirection and anisotropy, but also draws on the properties of multisolution and time-frequency localization inherited from wavelet transform. Contourlet transform can express contours sufficiently and capture more image directional information. But the downsamplers and upsamplers presented in both LP and DFB. Thus, it is not shift-invariant which causes pseudo-Gibbs phenomena around singularities. To overcome the shorting, nonsubsampled pyramid structure and nonsubsampled directional filter banks are employed in Nonsubsampled contourlet transform(NSCT). This papper has researched the image denoising methods based NSCT, the main contents are as follows:First, study the theory and composition of Contourlet transform and NSCT.Second, considering inter-scale and intra-scale dependency, in this paper, an image denoising method in NSCT domain by using locally adapt bivariate shrinkage algorithm is proposed. Simulation results and analysis indicate that the proposed algorithm obviously outperforms classical algorithms in both PSNR and visual quality, achieve better preservation of sharp details and directional information.At last, we analysis characteristics of SAR noise, the undecimated discrete wavelet transform (UDWT) is used to code homogeneous areas while NSCT is used to code edges areas f the SAR image. The segmentation between homogeneous areas and edges areas is done by using total variation (TV) segmentation. By combining the attributes of both transformations, it is possible to denoise SAR images better than the classical algorithms not only in value of ENL, STD and MR, also in human visual quality.