摘要：One of the biggest problems faced while analyzing digital
elevation models (DEMs), particularly DEMs that are produced using
photogrammetry, is to avoid pits and peaks in DEMs. Peaks and
pits, which are errors, are generated during the surface
generation process. DEM smoothening is an important preprocessing
step meant for removing these errors. This paper discusses two
linear DEM smoothening methods, Gaussian blurring and mean
smoothening, and two nonlinear DEM smoothening methods,
morphological smoothening and morphological smoothening by
reconstruction. The four methods are implemented on a
photogrammetrically generated DEM. The drainage network of the
resultant DEM is obtained using skeletonization by morphological
thinning, and the fractal dimension of the extracted network is
computed using the box dimension method. The fractal dimensions
are then compared to study the effects of the four smoothening
methods. The advantages of nonlinear DEM smoothening over linear
DEM smoothening are discussed. This study is useful in landscape
descriptions.
