文献类型：期刊文献

中文题名：基于小波变换的高分辨率快鸟遥感图像薄云去除

英文题名：Wavelet-based Cloud Removal from High-resolution Remote Sensing Data:An Experiment with QuickBird Imagery

作者：张波[1];季民河[2,1];沈琪[3]

机构：[1]华东师范大学地理信息科学教育部重点实验室,上海200062;[2]广西科学院广西-东盟海洋科学研究中心,南宁530007;[3]华东理工大学商学院,上海200237

年份：2011

卷号：33

期号：3

起止页码：38

中文期刊名：遥感信息

外文期刊名：Remote Sensing Information

收录：CSTPCD;;CSCD:【CSCD_E2011_2012】；

基金：广西科学基金(桂科回0639007)

语种：中文

中文关键词：遥感图像处理;去云;同态滤波;小波变换;QuickBird

外文关键词：cloud removal; homomorphic filtering; fourier transform; wavelet transform; QuickBird

摘要：薄云污染会影响遥感图像的正常判读和解译。传统去云方法对图像做傅里叶变换,对转为频率域后的图像进行同态高通滤波的整体处理,因而会在去云的同时对无云区域及图像边缘产生较大影响。本文利用小波变换将图像分解为若干频率特征不同的分量,仅仅针对表示薄云的低频近似分量进行同态滤波,最后通过小波重构得到去除薄云的图像。试验结果表明,小波变换使具有高频细节的地物信息免受滤波处理,但连续变化的低频地物信息仍会受到一定影响。
Thin cloud is considered annoying contamination on remote sensing imagery,as it can seriously affect the image analysis.Cloud removal is commonly performed with a homomorphic filter（e.g.the butter worth or exponential filter） applied to the fourier transformed frequency image.One drawback of the method is that it is unable to separate the low-frequency component from the high frequency in the filtering process,thus leading to information loss in the non-cloud region.This paper presents a homomorphic filtering based on wavelet transformation for cloud removal,which can avoid the above drawback.This procedure decomposes an image into low-frequency and high-frequency components,and then employs a high-pass filter to process the low-frequency（cloud） component.The filtered image is finally recovered from all the other components with this processed one through wavelet reconstruction.Test evaluations indicated that the filtered image resulting from the wavelet-based approach performed better than the image from the conventional fourier transformation in terms of changes in the mean,standard deviation,and entropy of image brightness values.