Multiresolution Analysis of DEMs: Error and Artifact Characterization

Digital elevation data is often affected by errors due to excessive interpolation or to interferometric artifacts. The local analysis of the DEM roughness allows both the detection of these errors and the segmentation of the elevation data for a better understanding of the existing geological structures. This analysis can be performed by means of fractal dimension estimators. We compare several fractal estimation methods and show that those based on the multiresolution (wavelet) data analysis yield the best results from the point of view of their segmentation capabilities. This is a natural conclusion considering that fractals and wavelets rely on many common concepts. As an example we show an application in which SAR intereferometric artifacts are revealed and the elevation data is separated in different roughness classes using fractal dimension measurements and an unsupervised clustering algorithm.