Efficient Multi-Band Texture Analysis for Remotely Sensed Data Interpretation in Urban Areas

Texture analysis is a long-standing and important problem in image-based urban characterization. A variety of approaches and methods have been proposed in the past to deal with urban texture segmentation and classification. However, texture characterization is particularly complex when the image data is composed of several spectral bands at different wavelengths, as in the case of remotely sensed hyperspectral images, in which hundreds of spectral bands are often available. Such images have two domains which can be analyzed: the spectral domain and the spatial domain. In this paper, we develop a joint spatial/spectral classification approach for hyperspectral imagery which is shown to perform effectively in highly complex urban environments. Experimental results are provided using a hyperspectral scene with extensive ground-truth, collected over the town of Pavia in Italy. To address the high computational requirements of the algorithm, we also develop a parallel implementation which is tested in this work using a massively parallel supercomputer at NASA's Goddard Space Flight Center in Maryland.