Empirical Quality Assessment: Effect of Resampling on Geometric and Radiometric Data Quality Using a Regionbased Approach

Registration and geocoding of remote sensing data and ground truth is necessary in numerous applications. An often applied technique consists in a resampling procedure regardless of well known disadavantages. Here quantitative results are presented to show the influence of geometric transformation and radiometric interpolation on the data with emphasis on the influence on a single object. The study was carried out with data from a DAEDALUS AADS 1268 ATM airborne multispectral scanner. Test measurements are based on morphologic and internal descriptors. It can be shown, that even for the nearest neighbor algorithm, the changes of the radiometric properties of an object in such a dataset are not negligible. The geometric reliability does not seem to be satisfying, even for the case of a cubic convolution algorithm.