Gray Scale Log Residual method for reflectance pattern analysis

Abstract

Importance of conversion from digital numbers of remote sensing data to physical units has been stressed these days. For instance, the Log Residual method was proposed to relate digital numbers to reflectance patterns. However, the apparent spectral pattern given by the Log Residual method is not similar to the real reflectance spectrum, if this method is applied to the which contain one dominant cover type. The Least Upper Bound Residual method, which uses a spectral pattern with the maximum fluctuation as a reference, was proposed to avoid this problem. In this paper, we expose the problems of these methods and propose Gray Scale Log Residual (GSLR) method instead.
The GSLR method, which is a new technique to convert digital numbers to "apparent" reflectance patterns, is a modification of the original Log Residual technique in the following aspect ; to use reference pixel data which presumably exhibit flat (gray) reflectance pattern, e.g. cloud, halite and alluvial fan, in the visible to short-wave-infrared regions.
In this study, the GSLR method is evaluated by using cumulus cloud as reference pixels. Although the reference does not indicate a completely flat (gray) reflectance pattern, the GSLR method is the best of all the techniques in conversion from original data to apparent reflectance patterns.