Complex Discrete Wavelet Transform-Based Image Denoising

Abstract

Dual tree complex discrete wavelet transform is implemented for denoising asan important image processing application. Two wavelet trees are used, onegenerating the real part of the wavelet coefficients tree and the other generating theimaginary part tree.A general computer program computing two dimensional dual tree complexwavelet transform is written using MatLab V.7.0. for a general (NxN) twodimensional signal.This paper introduces firstly a proposed method of computing one and twodimensionaldual tree complex wavelet transform .The proposed method reducesheavily processing time for decomposition of image keeping or overcoming thequality of reconstructed images. Also, the inverse procedures of all the abovetransform for multi- dimensional cases verified.Secondly, many techniques are implemented for denoising of gray scale image.A new threshold method is proposed and compared with the other thresholdingmethods. For hard thresholding, PSNR gives (13.548) value while the PSNR wasincreased in the proposed soft thresholding, it gives (14.1734) PSNR value whenthe noise variance is (20).Denoising schemes are tested on Peppers noise image to find its effect ondenoising application. The noisy version has SNR equals to (11.9373 dB), thedenoising image using WT has SNR equals to (17.4661 dB), the denoising imageusing SWT has SNR equals to (18.1459 dB), the denoising image using WPT hasSNR equals to (19.3640 dB), the denoising image using Complex DiscreteWavelet Transform has SNR equals to (21.9138 dB) using hard threshold and hasSNR equals to (22.1393 dB) using soft threshold. Matlab V.7.0 is used forsimulation.