A texture-tone analysis for automated landuse mapping with panchromatic images

Texture in digital data processing means the spatial distribution of tones of the pixels. A new texture analysis is presented based on extracting both spatial-tone and wave-form parameters from 3x3 and 5x5 pixels, referred to as Model I and Model II, for classifying individual pixels into one of the training sets or a reject category. The classifier used is the linear discriminant functions obtained from the inverse of the dispersion matrix of the texture-tone variables. A hit-rate of over 90% is obtained for the classification of general land-use types with panchromatic images.