结合有向图模型和改进Bandlet变换的图像去噪算法  

Image denoising algorithm based on digraph model and improved Bandlet transform

作　　者：郭勇 潘力 GUO Yong;PAN Li(North Sichuan College of Preschool Teacher Education,Guangyuan 628000,Sichuan Province,China;Zhengzhou University of Technology,Zhengzhou 450044,China)

机构地区：[1]川北幼儿师范高等专科学校,四川广元628000 [2]郑州工程技术学院,郑州450044

出　　处：《信息技术》2020年第12期22-27,32,共7页Information Technology

基　　金：河南省科技厅科技攻关计划项目(202002210346)。

摘　　要：针对图像中具有较多噪声点的问题,利用变换域建模的稀疏性和多尺度特性,提出了一种结合有向图模型和Bandlet变换的图像去噪算法。其首先将图像建模成有向图模型,然后在Bandlet变换的基础上,融入了基于图模型的图像几何表示,即图模型上进行Bandlet变换,提取特征值。其中,用一种新的非正交小波代替了传统的小波,提高了Bandlet变换的方向选择性。最后,通过小波变换和图像重建来实现图像的去噪功能。在两种图像数据集上的实验结果表明,该方法在信噪声比、CNR等指标上都具有明显的优势,具有良好的去噪性能。Aiming at the problem that there are many noise points in the image,an image denoising algorithm combining the digraph model and Bandlet transform is proposed by using the sparsity and multi-scale characteristics of transform domain modeling.Firstly,the image is modeled as a directed graph model,and then on the basis of Bandlet transform,the geometric representation of the image based on the graph model is integrated,that is,the Bandlet transform is carried out on the graph model to extract the feature value.Among them,a new non orthogonal wavelet is used to replace the traditional wavelet,which improves the direction selectivity of Bandlet transform.Finally,the denoising function of image is realized by wavelet transform and image reconstruction.The experimental results on two kinds of image data sets show that the method has obvious advantages in SNR,CNR and other indicators,and has good denoising performance.

关 键 词：图像去噪 变换域建模 有向图模型 改进Bandlet变换 非正交小波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]