一种改进的二阶偏微分去噪模型  

Improved Second Order Partial Differential Model for Image Denoising

作　　者：杨平先[1] 陈明举[1] 

机构地区：[1]四川理工学院自动化与电子信息学院,四川自贡643000

出　　处：《电视技术》2014年第23期38-40,共3页Video Engineering

基　　金：人工智能四川省重点实验室基金项目(2011RZY01;2012RYY08);四川省教育厅基金重点项目(10ZA135);四川高校科研创新团队建设计划资助项目(13TD0017);四川理工学院校级科研基金项目(2012KY13)

摘　　要：分析了基于二阶偏微分的扩散方程模型的基本原理,针对该模型在去噪的同时会产生阶梯效应的缺点,提出了一种基于图像结构信息的二阶偏微分去噪模型。在该模型中,在二阶偏微分的全变分模型的正则化项加入图像的切矢量和法矢量信息,并由此推导出相应的扩散方程,再对扩散方程的扩散强度因子进行修改。在实验中,将模型分别与基于二阶偏微分、四阶偏微分的全变分模型进行对比分析表明,实验结果证明该模型能有效地去除图像噪声,克服阶梯效应的产生,主观性能最优。Through discussing the characteristics of the second order partial differential equation for image noise removal. A novel total variational denoising model based the image structure is proposed to overcome staircase effect which is caused by second order partial differential equation. In this proposed TV model ,the tangents and normals directions to the isophotes of the original image is employed in regularization item and to obtain diffusion equation, then use gradient to correct the diffusion strength parameters. Compared to second order and fourth-order diffusion, the experimental results show that the proposed model can remove the noise diffusion, preserving more image details and avoid staircases, it has good performance in image denoising.

关 键 词：图像去噪 全变分模型 阶梯效应 图像结构 

分 类 号：TN911.73[电子电信—通信与信息系统]