COMBINING SPECTRAL AND TEXTURE DATA IN THE SEGMENTATION OF REMOTELY-SENSED IMAGES

Image segmentation is a method of defining discrete objects or classes of objects in images. Addition of a spatial attribute, i.e., image te xture, improves the segmentation process in most areas where there are differences in texture between classes in the image. Such areas inclu de sparsely vegetated areas and highly textured human-generated areas, such as the urban-suburban interface. A simple adaptive-window textur e program creates a texture channel useful in image segmentation. The segmentation algorithm is a multi-pass, pair-wise, region-growing algo rithm. The test sites include a simulated conifer forest, a natural ve getation area, and a mixed-use suburban area. The simulated image is e specially useful because polygon boundaries are unambiguous. Both the weighting of textural data relative to the spectral data, and the effe cts of the degree of segmentation are explored. The use of texture imp roves segmentations for most areas. It is apparent that the addition o f texture, at worst, has no influence on the accuracy of the segmentat ion, and can improve the accuracy in areas where the features of inter est exhibit differences in local variance. Results indicate that, for most uses, segmentation schemes should include both a minimum and maxi mum region size to insure the greatest accuracy.