基于Contourlet变换的高空间分辨率遥感影像检索研究  

Retrieval of high spatial resolution remotely sensed imagery based on Contourlet transform

作　　者：许锐[1,2] 曾艳芳[3] 

机构地区：[1]福州大学福建省空间信息工程研究中心,空间数据挖掘与信息共享教育部重点实验室,福州350002 [2]福建工程学院计算机与信息科学系,福州350108 [3]福建师范大学旅游学院,福州350108

出　　处：《测绘科学》2013年第4期116-118,共3页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(40871206);国家重点基础研究发展计划(973计划)项目子课题(2006CB708306);福建省教育厅科技研究项目(JB11116)

摘　　要：本文在高空间分辨率遥感影像纹理特征提取研究中引入Contourlet变换,将不同尺度、不同方向子带系数矩阵的均值、方差和能量作为纹理特征,并为分类能力强的特征量赋予较大的权值,体现相应子带分类能力的差异性。以QuickBird影像为样本数据、采用五叉树分解策略的纹理检索实证研究表明,基于Contourlet变换提取高空间分辨率遥感影像纹理特征信息的效果优于Gabor小波变换。Contourlet transform was introduced in the research of texture extraction from high spatial resolution remotely sensed imagery. The mean, variance, and energy of sub-band coefficient matrix in different scales and directions were taken as texture fea- tures. Bigger weights were assigned to the features with better discrimination quality. A case study of texture retrieval, in which Queik- Bird imagery were taken as sample data and the Quintuple tree decomposition strategy was employed, showed that the proposed method based on Contourlet transform would be superior to the one based on Gabor wavelet transform in terms of the result of texture feature in- formation extraction from high spatial resolution remotely sensed imagery.

关 键 词：CONTOURLET变换 五叉树分解 遥感影像 纹理检索 

分 类 号：P283.49[天文地球—地图制图学与地理信息工程]