基于L2范数的高斯混合模型空域去噪方法  

A Spatial Denoising Method Based on Gaussian Mixture Model and L2 Norm

作　　者：王彩玲[1] 孙昌霞[2] WANG Cailing;SUN Changxia(Department of Network Security, Henan Police College,Zhengzhou 450046,China;Computer Department of Information Management College, Henan Agricultural University, Zhengzhou 450046,China)

机构地区：[1]河南警察学院网络安全系,河南郑州450046 [2]河南农业大学信管学院计算机系,河南郑州450046

出　　处：《探测与控制学报》2021年第2期94-100,109,共8页Journal of Detection & Control

基　　金：河南省科技攻关项目资助(162102210109)。

摘　　要：针对图像像素间相似性衡量不准确而导致去噪性能下降的问题,提出了基于L2范数的高斯混合模型空域去噪方法。该方法将像素点周围像素的局部灰度信息的统计特性建模为高斯混合模型,结合高斯混合模型间的L2范数和像素位置的空间距离来定义像素点之间的相似性权值,以提高像素间相似性的度量的精确度。实验结果表明,利用高斯混合模型可更加准确地衡量像素间的相似性,同时,该去噪方法可提高图像的去噪效果,并能较好地保留图像的细节信息。To solve the problem that the image pixel similarity measurement was not accurate,which led to a degradation in the denoising performance,a spatial denoising method based on Gaussian mixture model and L2 norm was proposed.The statistical characteristics of local gray information of pixels around pixels were modeled as Gaussian mixture model,and the similarity weights between pixels were defined by combining the L2 norm between Gaussian mixture models and the spatial distance of pixel positions,so as to improve the accuracy of similarity measurement between pixels.The experimental results showed that the Gaussian mixture model could measure the similarity between pixels more accurately.Simultaneously,the denoising method could improve the image denoising effect,and could better retain the image details.

关 键 词：高斯混合模型 图像去噪 L2范数 像素信息 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]