Filtering of ASTER digital elevation models using mathematical morphology

Authors:
  • CFC. Fleming
    C. Fleming
  • JGJ. R. A. Giles
    J. R. A. Giles

Abstract
This paper presents results of research undertaken on the creation and filtering of digital elevation models (DEMs) from a stereo pair of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images. The raw DEM, created by automatic image matching, appears to be very noisy. Two types of irregularities can be observed. First, a random occurrence of small sinks and mounds with high amplitude is observed. Secondly, a more regular east–west-oriented pattern of noise is present. Many DEM-creation programs provide some editing tools to smooth out the irregularities, including some noise removal, smoothing and interpolation algorithms. However, the application of these algorithms has an important impact on the values of the parameters derived from the elevation, such as slope, aspect and curvature. In this study we propose a filtering algorithm based on morphological greyscale reconstruction in order to remove the sinks and mounds. This technique is very effective in mitigating the artefacts while preserving the remaining structures. For the regular pattern, a linear north–south-oriented low-pass filtering showed the best results. This approach was compared with a median filter and proved to be more effective in terms of both elevation and slope parameters.

This publication has 12 references indexed in Scilit: