A unifying approach to isotropic and anisotropic total variation denoising models
A unifying approach to isotropic and anisotropic total variation denoising models
复制标题
DOI：
10.1016/j.cam.2010.11.003
复制
发表时间：
2011-02-15
影响因子：
2.4
通讯作者：
Birkholz, Harald
中科院分区：
文献类型：
--
作者：
Birkholz, Harald
总变差最小化是一种成熟的数字图像恢复方法。它的隐式保存的边缘允许导出的各向异性模型的定性改善角落。本文简要介绍了各向异性模型，并对各向同性模型的数值进行了最新的分析。我们概括了两个有代表性的模型，从这两个分支的研究。这种提法导致一个一般的收敛算法和一个一般的高效算法，适用于这两种情况。从各向异性模型到各向同性设置的离散化的转移导致旋转不变性的改善。(C)2010 Elsevier B.V.保留所有权利。
Total variation minimisation is a well-established method for digital image restoration. Its implicit preservation of edges permits the derivation of anisotropic models for a qualitative improvement at corners. This paper is a synopsis of anisotropic models with state-of-the-art insights into the numerics of isotropic models. We generalise two representative models from both branches of research. This formulation leads to a general convergent algorithm and a general highly efficient algorithm which apply for both cases. A transfer of the discretisation from the anisotropic model to the isotropic setting results in an improvement of rotational invariance. (C) 2010 Elsevier B.V. All rights reserved.