摘要
针对现有去除乘性噪声的变分模型存在细节丢失和计算速度慢的问题,文中引入权重函数,在此基础上给出一种基于偏微分方程(PDE)的去除图像乘性噪声的变分模型。为了提高运算速度,在该模型中引入不精确的交替方向乘子算法(IADMM)。在算法中,引入辅助变量将原问题变为3个相关的子问题,然后分别对3个子问题求解。实验结果表明,模型有较好的去噪效果,能够较好地抑制图像中的"阶梯效应"。与梯度下降法相比,该算法处理过程快,极大地缩短了运算时间,并且保持了较好的去噪效果。
A weight function is introduced and a new variational model based on partial differential equa- tion (PDE) is proposed to solve the problem about the loss of detail and the slow calculation speed of the existing variational approaches to remove multiplieative noise. An inexact alternating direction method of muhiplier (IADMM) is introduced to improve processing speed in the model. Due to the introduction of the auxiliary variable, the primal problem is divided into three relevant subproblems in the algorithm, and then these subproblems are respectively solved. Experimental results show that the model has a good de- noising effect and restrains the "staircase effect". Compared with the gradient descent method, the algo- rithm is faster and can keep denoising effect better.
出处
《南京邮电大学学报(自然科学版)》
北大核心
2016年第3期74-78,共5页
Journal of Nanjing University of Posts and Telecommunications:Natural Science Edition
基金
国家自然科学基金(11301281)资助项目
关键词
图像去噪
伽马噪声
偏微分方程
不精确的ADMM(IADMM)算法
image denoising
Gamma noise
partial differential equation (PDE)
inexact alternating di-rection method of multiplier(IADMM) algorithm