Applicability of category decomposition for the fusion of multi-resolution data

Abstract

The authors tested the applicability of the category decomposition method based on the linear mixture model for the fusion of multiple-resolution satellite data such as Landsat-TM and NOAA-AVHRR. The goal of the application of this method is to estimate the mixing ratio of different categories within one pixel of the lower-resolution data using the classification result of the higher-resolution data, which is considered to be useful for the extrapolation of the information from the higher-resolution data over the wider coverage of the lower-resolution data. The authors tested the estimation accuracy by two kinds of decomposition methods, the maximum likelihood estimation and the minimum distance estimation and also by the multiple regression method. The experimental results showed that the most adequate estimation was obtained by the category decomposition based on the minimum distance estimation.