Variational total curvature model for multiplicative noise removal

• DOI: 10.1049/iet-cvi.2017.0332
• Publication Date: 2018-01-19T13:23:05Z
• Authors: Xinli Xu; Teng Yu; Xinmei Xu; Guojia Hou; Ryan Wen Liu; Huizhu Pan
• Keywords: Hessian matrix; Smoothness; Multiplicative noise

The multiplicative noise removal problem has received considerable attention recently. To solve this problem, various variational models have been proposed, which minimise an energy functional composed of the data term and regularisation term. Regarding term, a first‐order model is frequently used to remove noise, may cause staircase effect loss contrast in output image. In study, authors use second‐order model, total curvature (TC), above problem. TC benefit removing maintaining image edges, contrasts corners. augmented Lagrange method utilised proposed by introducing auxiliary variables, multipliers using alternating optimisation strategy. each loop optimisation, fast Fourier transform, generalised soft threshold formulas, projection gradient descent are integrated effectively. experimental results show that can effectively preserve smoothness, via comparison with (total variation Perona–Malik regularisation). Furthermore, better than another based on bounded Hessian preserving corner.