Noise classification and automatic restoration system using non-local regularization frameworks

ABSTRACT

Medical, satellite or microscopic images differ in the imaging techniques used, hence their underlying noise distribution also are different. Most of the restoration methods including regularization models make prior assumptions about the noise to perform an efficient restoration. Here we propose a system that estimates and classifies the noise into different distributions by extracting the relevant features. The system provides information about the noise distribution and then it gets directed into the restoration module where an appropriate regularization method (based on the non-local framework) has been employed to provide an efficient restoration of the data. We have effectively addressed the distortion due to data-dependent noise distributions such as Poisson and Gamma along with data uncorrelated Gaussian noise. The studies have shown a 97.7% accuracy in classifying noise in the test data. Moreover, the system also shows the capability to cater to other popular noise distributions such as Rayleigh, Chi, etc.

KEYWORDS:

• Noise estimation
• noise classification
• image restoration
• denoising
• non-local total variation
• regularization

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

P. Jidesh http://orcid.org/0000-0001-9448-1906

Notes on contributors

I. P. Febin received her M.Tech degree in Computer Science with specialization in image processing from Cochin University, India, in 2016. Presently she is working as a Junior Research Fellow in the Department of Mathematical and Computational Sciences, National Institute of Karnataka, India, for the project funded by the Science and Engineering Research Board, Govt. of India. Her research interests are image processing, pattern recognition, and computer vision.

P. Jidesh received Ph.D. from the Department of Mathematical and Computational Sciences, National Institute of Technology, Karnataka, Surathkal, India. Since January 2009 he is with the Department, where he is currently an Assistant Professor. Dr Jidesh has published several papers in reputed International Journals and Conferences. His areas of research interest include Mathematical imaging, Graph image processing, and Data compression.

A. A. Bini completed her Ph.D. in 2015 from the Department of Electronics and Communication Engineering, National Institute of Technology, Karnataka, India. She is presently working as an assistant professor at Indian Institute of Information Technology, Kottayam. Dr Bini has published several papers in reputed journals. Her areas of research interest include signal processing, image processing, communication engineering, etc.

Additional information

Funding

The authors (I. P. Febin and P. Jidesh) would like to thank the Science and Engineering Research Board, Government of India for providing the financial support under the Project Grant No. ECR/2017/000230.