基于非线性扩散方程的改进全变分去噪模型  

Improved Total Variation Denoising Model Based on Nonlinear Diffusion Equation

作　　者：郗金洋 段锦[1] 郭红芮 韩学辉[1] 陈宇[1] XI Jin-yang;DUAN Jin;GUO Hong-rui;HAN Xue-hui;CHEN Yu(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022)

机构地区：[1]长春理工大学电子信息工程学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2021年第3期90-95,共6页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金(61890963)。

摘　　要：针对传统的全变分(TV)模型因参数选择敏感,导致去噪图像容易在平滑区域产生"阶梯效应"或者虚假边缘的情况,提出一种基于非线性扩散方程的改进全变分去噪算法。在传统的TV算法基础上,提出了一种针对参数的自适应迭代函数,结合P-M算法的非线性扩散方程,使本算法在迭代初期可以看作各向同性去噪模型,有效去除"阶梯效应",随着迭代次数的增加,此模型为各向异性去噪模型,在去除噪声的同时,有效保护图像的边缘细节。实验结果表明,去噪过程中,该算法在扩散系数和自适应迭代函数的共同作用下,消除了阶梯效应和虚假边缘,相比传统TV算法提升了图像3 dB的峰值信噪比(PSRN)和视觉效果。基本满足图像预处理要求。In view of the fact that the traditional total variation(TV)model is sensitive to parameter selection,which leads to"ladder effect"or false edge in smooth area,an improved total variation denoising algorithm based on nonlinear diffusion equation is proposed in this paper.Based on the traditional TV algorithm,an adaptive iterative function for parameters is proposed.Based on the traditional TV algorithm,an adaptive iterative function aiming at different parameters is proposed.Combined with the nonlinear diffusion equation of P-M algorithm,this algorithm can be regarded as an isotropic denoising model at the beginning of iteration,which can effectively remove the"ladder effect".With the increase of the iteration number,the anisotropic denoising model can protect the edge details of the image effectively while removing the noise.Experimental results show that in the process of denoising,the algorithm eliminates ladder effect and false edge under the joint action of diffusion coefficient and adaptive iterative function,and improves the peak signal-to-noise ratio(PSRN)and visual effect of 3 dB image compared with traditional TV algorithm.It basically meets the requirement of image preprocessing.

关 键 词：TV模型 图像去噪 非线性扩散 自适应 各项异性 图像梯度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]