作者：刘春国

第一作者单位：河南理工大学测绘与国土信息工程学院

摘要：针对LANDSAT-5 TM多光谱图像存在散粒噪声问题,提出一种基于异常探测和光谱归一化处理来检测及去除噪声的算法。视散粒噪声为强光谱异常,基于RX算子探测像元异常,根据异常分布统计特征确定含散粒噪声的像元;再以含散粒噪声的像元为中心,利用邻域内像元各分量上的光谱归一化分值,确定散粒噪声所在波段。利用邻域内具有最相似光谱的像元对应分量替换散粒噪声所在的像元分量。实验表明,该算法能够有效去除TM图像中的散粒噪声。

Abstract：Shot noise is existed in the LANDSAT-5 TM multispectral image.In this study,an algorithm based on anomaly detection is proposed to detect and remove the noises.Firstly,an RX anomaly detection operator is used to detect spectral anomalies of pixels including strong anomalies caused by the noises.Secondly,the pixels containing a shot noise component are determined according to the statistical characteristics of the abnormal distribution.Thirdly,a neighborhood centered at pixel containing a shot noise component is constructed.The normalized score of each component of the pixels within the neighborhood is calculated,and then is used to determine which component of center pixel is shot noise.Finally,the shot noise component of center pixel is substituted by the corresponding component of pixel within the neighborhood which has the most similar spectral features.The experimental results show that the algorithm effectively identify and remove the shot noise in the TM image.

基金：中国地质调查局地质调查项目(1212011120673)；

关键词：LANDSAT-5 TM多光谱图像;散粒噪声;异常检测;RX算子;光谱归一化;噪声去除;

DOI：10.16186/j.cnki.1673-9787.2018.01.010

分类号：TP391.41