The effects of terrain illumination correction on automatic classification

In Remote Sensing analysis it was pointed out that the reflectance of many land covers is influenced by solar illumination, that differs in dependence of the shape of the terrain. These terrain illumination effects can affect automatic classification methods, due to the fact that similar land covers produce different reflectance values. In the present study we compared some radiometric image correction methods applied to reduce differential illumination effects to an IKONOS image dataset. In literature Cosine Correction (1) based on the assumption of an isotropic (Lambertian) reflectance, is widely argued. However several authors showed that this method overcorrects the images (2, 3, 4) mainly in faintly illuminated areas. In our research we applied the C-correction method (1, 3) and a modified version of the Normalization method (5). Finally we performed an automatic classification to assess if the proposed DEM based altimetry effect correction improves the results.