Geostatistical description of texture on an aerial photograph for discriminating classes of land cover

The mean square-root pair difference between the DN values of pixels separated by a particular spatial interval (lag) in a digital image is a measure of spatial variability related to the semi-variance. It is shown that this statistic, when evaluated in a local window for several lags differing in length and direction, can make an important contribution to the discrimination of classes of land cover. Square-root pair differences are, for this purpose, statistically preferable to semi-variances. It is also shown that a set of square-root pair differences measures different components of image texture, not all of which may be distinguished by simpler textural variables.