Context-sensitive Filtering of Terrain Data based on Multi Scale Analysis

Abstract

This paper addresses triangle count reduction for regular terrain triangulations. Supported by additional information available from cadastre data, non-linear low pass filtering based on wavelet coefficients is applied. The aim is to generate a simplified view of a city model's terrain so that still landmarks can be easily recognized. Height data, that are not critical for visual impressions, are compressed with lower resolution than height data belonging to landmarks. Results are presented for the German city of Krefeld.