Practical Illumination Correction of Satellite Imagery in Consideration of Anisotropic Skylight Component

In satellite images over rugged terrain, we can find some topographic effect entangled with atmospheric effects [1]. Without correcting these effects, it is difficult to evaluate the satellite images for land cover classification or multi-temporal monitoring of natural environment. Iikura[2] derived a unified simple correction formula (DN − B)/(cosβ + C) from a physical model of signal detected by sensor on board. Although either of the parameters B or C is ignored in the practical illumnation correction such as C correction method [3], and modified cosine method [4], both methods seem to work well for satellite images with high sun elevation by estimating one of these parameters from regression of digital nunbers (DNs) to cosine of solar incident angle (β). In this paper, we will estimate these parameters based on the radiative transfer equation (6s) [5] and exact calculation of topographical parameters such as sky view factor. It will be shown that the parameter C becomes smaller than we expected when we take account of an anisotropic skylight component which is propotional to cosβ.