De-Noising of MRI Brain Tumor Image using Deep Convolutional Neural Network

Abstract

Medical images must be introduced to specialists or doctors with high precision for the diagnosis of critical diseases like a brain tumor. Presently, a lot of research is going on in the area of image denoising and found to be an emerging one. In this paper, a pretrained feed-forward denoising using convolutional neural networks (DnCNNs) is considered and found to be better than conventional filters used. The results are compared with the Gaussian, adaptive, bilateral and guided filters to reduce the noise from the image. Different noises like Salt and pepper, Gaussian, Poisson, and Speckle noise with variance of 0.05 are used to corrupt the input image and result in the noisy image. Performance metrics Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index of Image (SSIM) of the denoised image are calculated and compared across all filters and noises. The proposed DnCNN method has high PSNR and SSIM for all noises in terms of result performance.