Urban area detection in SPOT images using multi-scale technique and map knowledge

This paper presents a novel approach to urban area detection in SPOT images. A wavelet transform is first applied to the SPOT image, which results in two sets of multi-scale representations of the image: a high frequency information portion and a low frequency information portion. Knowledge from geographic map is then used as a guide to explore the low frequency information portion which favors detection of region information in the image. Urban areas are detected by a region growing method using a hysteresis technique