Efficient Empirical Reflectance Retrieval in Urban Environments

In a complex urban environment, images acquired with very high spatial resolution are often plagued with shadows. Shadows distort the spectral features of materials, possibly crippling many image-based applications such as visualization, anomaly detection and classification. Moreover, it is very difficult to perform accurate atmospheric compensation when in shadow due to the complex interaction of radiative components. This paper proposes a novel empirical atmospheric compensation method that is applicable to both sunlit and shadowed regions, and does not require the use of 3-D geometrical data nor Digital Elevation Models (DEM). The resulting reflectance images were then tested for the purposes of visualization, anomaly detection and classification.