Image Denoising by Discrete Wavelet Transform with Edge Preservation

Abstract

Digital images are a big part of today's life and science. It is important to have a good quality images which is not always a case due to the different reasons. One of the common problems with digital images is presence of the various types of noise. Removing noise from digital images is an important research field widely studied in the past decades. In this paper, we combined three successful methods applied in the wavelet domain with the aim to improve the quality of the denosining. The discrete wavelet transformation was used to enable image processing in frequency domain. In order to remove noise, soft thresholding technique was combined with the median filter. To preserve the image sharpness, edge coefficients were kept and not affected by the denoising process. The proposed method was tested on four standard benchmark images. In the comparison to other methods from literature and in term of peak-signal-to-noise-ratio the proposed method achieved better results. Based on the structure similarity index measure, we can conclude that the proposed method is efficient for removing Gaussian noise.