A Novel Approach for Noise Removal from Magnetic Resonance Images using Averaging Reconstructed Images

Abstract

The objective of this paper is to provide a novel approach for denoising the image. Image denoising is an important image processing task. Image denoising refers to the recovery of image that has been contaminated by additive white Gaussian noise. To clear the noise several denoising methods are used. The novel denoising approach used here is based on AVeraging REConstructed Images (AVREC). The reconstruction is used to reduce noise while preserving almost all useful image information. The averaging of the reconstructed images further reduces reconstruction errors and noise. This approach first divides the spectrum of the noisy image into different parts. The reconstructed image is obtained from every such partial spectrum using a 2-D singularity function analysis model. Finally, the denoising is achieved through averaging all of the reconstructed images because each of the reconstructed images is expressed as the sum of the same noise-free image and a different smaller noise image. Magnetic Resonance Images are used as the experimental images. The performance of the AVREC method is compared with the bilateral filter and median filter. Also, the novel image denoising scheme is applied for different type of noises such as Poisson noise and Speckle noise. The experimental results show that the proposed approach performs well than the bilateral filter and the median filter.