Manuscript Title:

Denoising of MR images Using Reformed Structural Loss


A. Aaisha Nazleem, Dr. S. S. Sujatha

DOI Number:


Published : 2021-04-10

About the author(s)

1. A. Aaisha Nazleem - Research Scholar, Reg No (17233152162001) S.T. Hindu College, Nagarcoil.
2. Dr. S. S. Sujatha - Associate Professor, Department of Computer Science S.T. Hindu College, Nagarcoil.

Full Text : PDF


Noise reduction or denoising of images will help in getting true images from noisy images.The differentiations of noise from other part of the images are difficult because edge and texture are also having high frequency as like noise. This work addresses the issue of denoising in Magnetic Resonance Imaging (MRI). This work proposes a new technique known as Reformed Structural Loss based 3D Multi-scale Deep Neural Network (RSLMDNN-3D)which is a variation of Generative Adversarial Network (GAN). A generator and a discriminator circuit help the work to reduce the noise. In order to preserve more structural information, a reformed model of multi scale 3D CNN model is proposed as Generator of this GAN framework. The proposed work gives good results when compared to Wasserstein Generative Adversarial Network (WGAN) and CNN based approach.


Denoising of Mr images Using Reformed Structural Loss