摘要
针对经典全变差模型在进行椒盐去噪时不能有效保留图像边缘信息的问题,提出一种基于L1范数和自适应全变差正则化的椒盐噪声图像去噪方法.该方法在全变差和图像去噪模型的基础上构建了显式椒盐去噪模型,利用噪声像素的平均值计算自适应正则化参数,并有效保留图像边缘信息,使用原始对偶梯度算法求解显示模型,使所求得的数值解更加接近原始图像.实验结果表明:与其他方法相比,本文方法在PSNR和SSIM方面均优于对比的方法,可以有效去除高密度椒盐噪声.
To solve the problem that the classical total variation model can't effectively preserve the image edge information when performing salt and pepper denoising,an image denoising method based on L1 norm and adaptive total variation regularization for pepper and salt noise has been proposed.In this method,an explicit salt and pepper denoising model has been constructed on the basis of total variation and image denoising model,and the adaptive regularization parameters been calculated by means of the average value of noise pixels to effectively retain the image edge information.Finally,the original dual gradient algorithm has been used to solve the display model,which makes the numerical solution closer to the original image.The experimental results show that,compared with other methods,this method is better than the contrast method in PSNR and SSIM,and can effectively remove the high-density salt and pepper noise.
作者
何明
HE Ming(Institute of Construction Engineering and Art Design, Chongqing Industry Polytechnic College, Chongqing 401120, China)
出处
《西南师范大学学报(自然科学版)》
CAS
2021年第5期115-120,共6页
Journal of Southwest China Normal University(Natural Science Edition)
基金
重庆市社会科学规划项目(2017YBYS108)
重庆工业职业技术学院校级重点项目(GZY201709-2B).
关键词
椒盐噪声
自适应图像去噪
原始对偶梯度
全变差正则化
图像复原
salt and pepper noise
adaptive image denoising
primal dual gradient
total variation regularization
image restoration