基于K均值的超分辨率图像噪声识别方法  

Noise Recognition Method for Super Resolution Image Based on K-means

作　　者：窦国贤 周伟 喻成琛 郭力旋 DOU Guoxian;ZHOU Wei;YU Chengchen;GUO Lixuan(Anhui Jiyuan Software Co.,Ltd.,Hefei 230088,China)

机构地区：[1]安徽继远软件有限公司,合肥230088

出　　处：《自动化与仪表》2023年第9期68-72,77,共6页Automation & Instrumentation

基　　金：国网信通产业集团两级协同研发项目(546818220016)。

摘　　要：现有的大部分噪声识别算法都是基于图像中只存在单一类型高强度噪声的假设,很少有研究者对于常见的低强度混合噪声进行研究,该文研究基于K均值的超分辨率图像噪声识别方法。采用加权均值算法灰度化处理超分辨率图像并提取出图像特征。通过粒子群算法建立聚类中心求解函数,自适应调整粒子惯性因子、学习因子及其适应度程度,寻找出全局最优聚类中心,结合K均值进行图像特征分类,确定具有噪声特征的类别,得到超分辨率图像噪声识别结果。实验结果表明,所提方法能够有效识别出各种类型超分辨率图像噪声,保证识别效果与速度。Most existing noise recognition algorithms are based on the assumption that there is only a single type of high intensity noise in the image,and few researchers have studied common low intensity mixed noise.This article studies a super-resolution image noise recognition method based on K-means.Using the weighted mean algorithm to grayize super-resolution images and extract image features.Establish a clustering center solving function using particle swarm optimization algorithm,adaptively adjust particle inertia factor,learning factor,and their fitness level,find the global optimal clustering center,combine K-means for image feature classification,determine the category with noise features,and obtain super-resolution image noise recognition results.The experimental results show that the proposed method can effectively identify various types of super-resolution image noise,ensuring recognition effectiveness and speed.

关 键 词：K均值 聚类中心 超分辨率图像 噪声识别 特征分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]