Comparison of neural networks for suppression of multiplicative noise in images

The paper compares several neural network (NN) architectures for suppression of multiplicative noise. The images may contain sharp boundaries and large homogeneous areas. Convolutional and fully connected networks are investigated. It is shown that different architectures require significantly diffe...

Full description

Saved in:
Bibliographic Details
Published inKompʹûternaâ optika Vol. 48; no. 3; pp. 425 - 431
Main Authors Pavlov, V.A., Belov, A.A., Nguen, V.T., Jovanovski, N., Ovsyannikova, A.S.
Format Journal Article
LanguageEnglish
Published Samara National Research University 01.06.2024
Subjects
image processing
neural network
noise reduction
radar image
sar
speckle noise
Online AccessGet full text

Cover

More Information
Summary:The paper compares several neural network (NN) architectures for suppression of multiplicative noise. The images may contain sharp boundaries and large homogeneous areas. Convolutional and fully connected networks are investigated. It is shown that different architectures require significantly different amount of training data to reach the same noise suppression quality. Examples of NN requiring lower amounts of training data are presented.
ISSN:0134-2452
2412-6179
DOI:10.18287/2412-6179-CO-1400