Deep Neural Network-based Methods for Brain Image De-noising: A Short Comparison

Abstract

Various types of noise may affect the visual quality of images during capturing and transmitting procedures. Finding a proper technique to remove the possible noise and improve both quantitative and qualitative results is always considered as one of the most important and challenging pre -processing tasks in image and signal processing. In this paper, we made a short comparison between two well-known approaches called thresholding neural network (TNN) and deep neural network (DNN) based methods for image de-noising. De-noising results of TNNs, Dn-CNNs, Flashlight CNN (FLCNN) and Diamond denoising networks (DmDN) have been compared with each other. In this regard, several experiments have been performed in terms of Peak S ignal to Noise Ratio (PS NR) to validate the performance analysis of various de-noising methods. The analysis indicates that DmDNs perform better than other learning-based algorithms for de -noising brain MR images. DmDN achieved a PS NR value of 29.85 dB, 30.74 dB, 29.15 dB, and 29.45 dB for denoising MR image 1, MR image 2, MR image 3 and MR Image 4, respectively for a standard deviation of 15.