摘要
该文运用LANDSATTM数据对三种地统计学纹理量测方法用于图像分类的性能进行了比较和分析。研究发现,基于绝对值变差函数的纹理具有更好的性能。经典的变差函数和基于方根的变差函数的性能依赖于提取纹理时所用的窗口大小。
The performance of three geostatistical texture measures, variogram, madogram and rodogram, for multispectral image classification is evaluated using Landsat TM data for discrimination of different lithological units. Madogram is found to have better performance for image classification than both variogram and rodogram. The performance of variogram and rodogram depends on the window size at which texture is extracted.
出处
《地理与地理信息科学》
CSSCI
CSCD
北大核心
2003年第4期89-92,共4页
Geography and Geo-Information Science
基金
教育部留学回国人员科研启动基金项目
