首页|高斯曲率与波域LMS算法相结合的图像去噪扩散模型

高斯曲率与波域LMS算法相结合的图像去噪扩散模型

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
本文在充分研究传统PM模型的基础上,针对传统模型在模糊边缘细节等信息的不足,先利用图像的几何属性将高斯曲率作为检测算子引入到扩散模型中,将它作为扩散系数来保护边缘控制扩散,从而建立基于高斯曲率的图像去噪模型.考虑到噪声和图像的重要特征都集中在图像的高频部分,再采用小波变换进行小波分解,提取图像的高频部分,在小波域中运用最小均方误差(LMS)算法设计自适应阈值,进一步控制上述新扩散模型的扩散强度,提升去噪效果,建立基于高斯曲率与最小均方误差法的波域PM改进模型,最后将低频部分和经过新模型处理的高频部分进行小波重构,得到最终的去噪图像.实验结果表明,本文方法不仅能够有效去除图像噪声,同时也提升了对重要信息的保护.
Image Denoising Diffusion Model Combining Gaussian Curvature and LMS Algorithm in Wave Domain
In this paper,on the basis of fully studying the anisotropic diffusion model(PM model),ai-ming at the shortcomings of the traditional model in fuzzy edge details and other information,the geometric properties of the image are firstly used to introduce the Gaussian curvature into the diffusion model as the detec-tion operator,and it is used as the diffusion coefficient to protect the edge control diffusion,so as to establish the image denoising model based on Gaussian curvature.Considering that noise and important features of the image are concentrated in the high frequency part of the image,the wavelet transform is used for wavelet de-composition to extract the high frequency part of the image,and the least mean square error algorithm(LMS al-gorithm)is used in the wavelet domain to de-sign an adaptive threshold to further control the diffusion intensity of the new diffusion model and improve the denoising effect.A PM model of wavelet domain denoising based on Gaussian curvature and least mean square error algorithm is established.Finally,the low frequency part and the high frequency part processed by the new model are reconstructed by wavelet,and the final denoising image is obtained.Experimental results show that the new method can not only effectively remove image noise,but also improve the protection of important information.

image denoisingPM diffusion modelwavelet transformGaussian curvatureleast mean square

吴静、邵文莎、祝珊珊、周先春

展开 >

江苏开放大学,江苏省终身教育学分银行管理中心,南京,210036

南京信息工程大学人工智能学院,南京,210044

图像去噪 PM扩散模型 小波变换 高斯曲率 最小均方差

国家自然科学基金项目2023年度江苏省高校哲学社会科学研究一般项目2023年度江苏省社科应用研究精品工程社会教育(社科普及)专项项目2023年江苏开放大学校级教学改革研究项目

N0.620712382023SJYB079123SJB-1223-QN-21

2024

信息化研究
江苏省电子学会

信息化研究

影响因子：0.218
ISSN：1674-4888
年,卷(期)：2024.50(1)
  • 22