Academy & Industry Research Collaboration Center (AIRCC)

Volume 9, Number 2, February 2019

Effectiveness of U-Net in Denoising RGB Images


Rina Komatsu and Tad Gonsalves, Sophia University, Japan


Digital images often contain “noise” which takes away their clarity and sharpness. Most of the existing denoising algorithms do not offer the best solution because there are difficulties such as removing strong noise while leaving the features and other details of the image intact. Faced with the problem of denoising, we tried solving it with a Convolutional Neural Network architecture called the “U-Net”. This paper deals with the training of a U-Net to remove 3 different kinds of noise: Gaussian, Blockiness, and Camera shake. Our results indicate the effectiveness of U-Net in denoising images while leaving their features and other details intact.


Deep Learning, Image Processing, Denoising, Convolutional Neural Network, U-Net.