Multi-core Median Redescending M-Estimator for Impulsive Denoising in Color Images

• Published in: Pattern Recognition Vol. 12725; pp. 261 - 271
• Main Authors: Mújica-Vargas, Dante, Rendón-Castro, Arturo, Matuz-Cruz, Manuel, Garcia-Aquino, Christian
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2021; Springer International Publishing
• Series: Lecture Notes in Computer Science
• ISBN: 3030770036; 9783030770037
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-77004-4_25

In this paper, to reduce impulsive noise in color images we propose an extension of the Median Redescending M-Estimator. For that purpose, a multitasking approach was developed such as a multi-core processing in order to reduce in parallel the noise on R, G and B color channels. With this paradigm, an acceleration up to three times can be guaranteed compared to the sequential paradigm, while having the ability to reduce corrupted data up to densities of 80%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$80\%$$\end{document} of fixed-value and 40%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$40\%$$\end{document} of random-value impulsive noises, guaranteeing the preservation of edges. The effectiveness of our proposal is verified by quantitative and qualitative results.