Abstract:The texture is one of the important features of remote sensing images. In this paper, the image texture was extracted from Landsat TM data by means of semivariogram of logarithms, one of the geostatistic functions, and added to multispectral lithological classification. Different window sizes were used to extract textural information. The results of image classification show that the classification based on spectral data and geostatistical textural information can produce much higher overall accuracy than that based merely on spectral data. Moreover, for lithological discrimination based on multispectral data, the larger the window size for texture extraction is, the more accurate the classification result will become. In practice, however, other factors, such as the boundary effect and the accuracy of some important lithological units, need to be considered in choosing an appropriate window size.

THE APPLICATION OF GEOSTATISTICAL IMAGE TEXTURE TO REMOTE SENSING LITHOLOGICAL CLASSIFICATION