Title:	CHOICE OF THE REGULARIZATION PARAMETER FOR TOTAL VARIATION IMAGE DENOISING USING NO-REFERENCE METRIC
Author(s):	Nikolay Mamaev, Andrey Krylov and Dmitry Yurin
ISBN:	978-989-8533-79-1
Editors:	Katherine Blashki and Yingcai Xiao
Year:	2018
Edition:	Single
Keywords:	Total Variation Denoising, Image Denoising Quality Assessment, Mutual Information
Type:	Full Paper
First Page:	253
Last Page:	260
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	An automatic method for total variation denoising parameter calculation has been proposed. Parameter optimization is performed in the ridge areas of the difference image between original noisy and filtered images (so-called method noise image). Appearance of regular components on method noise is controlled using mutual information closely connected with conditional entropy. Images corrupted with Gaussian-like noise with small correlation radius are considered. Hessian matrix eigenvalues analysis is used for estimation of sizes and directions of image characteristic details. FISTA optimizer was used for the minimization of the regularizing total variation functional. Images with added controlled Gaussian noise from retinal DRIVE and general image BSDS500 databases, and biomedical images of nuclei of HL60 cell line were used for testing.
	Go Back