Using anisotropic bivariate threshold function for image denoising in NSCT domain

Abstract

In this paper, a more stable solving method of anisotropic bivariate Laplacian distribution function and corresponding threshold function is derived from the model using Bayesian estimation theory and extended to the non subsampled contourlet(NSCT) domain. A novel Non-Subsampled Contourlet Transform based on anisotropic bivariate threshold function (ABNSCT) for image denoising has been proposed. Such algorithms use anisotropic property of the variances of NSCT coefficients in different scales of natural images and a maximum a posteriori (MAP) relies on the conjecture that the NSCT coefficients and parameters locally vary with local marginal variance estimation. The simulation results indicate that the proposed method can remove Gaussian white noise effectively over a wide range of noise variance, improve the peak signal-to-noise ratio of the image, and keep better visual result in edges information reservation as well. © 2012 Springer-Verlag.