Improved image denoising based on Haar wavelet transform

Abstract

The wavelet decomposing levels and the selection of the thresholding function affect the performance of image denoising using the wavelet thresholding method. In this paper, a new method to identify the wavelet decomposing levels using the 2D Haar wavelet thresholding method is presented to denoise an image. It uses the standard deviation values of the sub-bands to find out if the signal energy is strong or weak in the high frequency sub-bands after the 2D Haar wavelet transform. In addition, a new thresholding function is proposed which achieves better denoising performance in terms of peak signal-to-noise ratio (PSNR) and mean squared error (MSE) than the soft thresholding method. Especially, at high noise levels, the proposed new thresholding method outperforms hard thresholding, soft thresholding and semi-soft thresholding methods.