基于TV模型对含乘性噪声图像的去噪算法1).pdf

与 TV A lgorithm to RemoveMu ltiplicative Noise Based on TV M ethod * 何 磊 潘振宽 魏伟波 H E L ei PAN Zhen- kuan WEI Wei- bo do:i 10. 3 6 /.j issn. 1672- 528. 200 . 03. 013 在针对乘性噪声的TV 模型的基础上, 将原本针对加性噪声的典型模型引入到乘性噪声模型中去在 行数值差分计算时,与常用的显式和半隐式差分实现方法不同, 本文采用备份迭代矩阵的方法来实现显 式差分, 该方法可使模型在去除噪声的同时, 能够保持图像的对比度不变在此基础上, 通过在正则化 项中加入常数, 使得模型在使用显式差分实现时稳定性大为提高, 从而简化了模型实现的难度, 提高了 算法的速度, 并且还一定程度上缓解了阶梯效应的出现试验结果表明, 改 的算法效果明显 TV模型 乘性噪声 变分方法 对比度 阶梯效应 显式差分 A bstract TV m ethod of removing mu ltiplicative noise is mi proved by add ing typical model of rem oving ad d itive noise. D ifferent from comm on exp licit scheme or sem i- mi plicit schem e, this paper mi plem ent exp licit difference scheme by back ing up the iterativem atrix. It can preserve mi age contrastw hile rem oving noise. Anoth er side, by add ing constant C in regularized term, the stability of exp licit d ifference schem e is m ore mi proved. It predigest mi plem ention of m odel and mi prove algorithm speed and reduce staircase effect. The valid ity of algo rithm s are proved by expermi ental resu lts. K eyw ord s TV m ethod M ultip licative noise Variational m ethods Contrast Staircase effect Exp licit difference schem e ( 3): 1 u u u x y