A modified convex variational model for multiplicative noise removal

作者：Liu, Min; Fan, Qibin^*

来源：Journal of Visual Communication and Image Representation, 2016, 36: 187-198.

DOI：10.1016/j.jvcir.2016.01.014

• Multiplicative noise
• Convexity
• Variational model
• Primal-dual methods
• Linearization
• I-divergence
• Statistical property
• Regularization

In this paper, a convex variational model for multiplicative noise removal is studied. Accelerating primal dual method and proximal linearized alternating direction method are also discussed. An improved primal-dual method is proposed. Algorithms above produce more desired results than primal-dual algorithm when we solve the convex variational model. Inspired by the statistical property of the Gamma multiplicative noise and I-divergence, a modified convex variational model is proposed, for which the uniqueness of solution is also provided. Moreover, the property of the solution is presented. Without inner iterations, primal-dual method is efficient to the modified model, and running time can be reduced dramatically also with good restoration. When we set parameter alpha to 0, the convex variational model we proposed turns into the model in Steidl and Teuber (2010). By altering alpha, our model can be used for different noise level.

• 出版日期2016-4
• 单位武汉大学