This
page contains animations that show some experiments performed with our two
anisotropic denoising algorithms, namely the anisotropic
smoothing (based on the anisotropic mean curvature flow) and
the prescribed smoothing (based on the prescribed mean
curvature flow).
The images of each
animation are consecutive snapshots. There was no post-processing
neither of the images nor of the animations.
Prescribed
smoothing of the dragon The
prescribed mean curvature flow is used to filter the
dragon corrupted by noise. The features of the surface are
preserved while the noise is removed.
Anisotropic
filtering of non-linear surface features A surface
with non-linear features and feature lines with varying curvature is
denoised with the prescribed mean curvature flow. The features are correctly preserved during the
smoothing process.
Denoising
of a surface with a vanishing ridge The fandisk is a standard model of a surface with
curved sharp edges and a vanishing ridge. The prescribed
smoothing correctly preserves the features while the anisotropic
smoothing flattens the ridge.
Anisotropic
smoothing of the Venus head Application
of anisotropic mean curvature flow to a high resolution model corrupted by noise.
Details of the surface are preserved while the noise is removed.
Prescribed
smoothing of the bone The
prescribed mean curvature flow is used to the smooth the
surface of a bone corrupted by noise. The features of the surface are
preserved while the noise is removed.
Prescribed
smoothing of the dragon The
prescribed mean curvature flow is used to the filter the
surface of the armadillo corrupted by noise. The features of the surface are
preserved while the noise is removed.
Comparison of
the anisotropic and the prescribed smoothing The
ring is heavily deformed by the anisotropic flow, because the curved parts
shrink while the flat parts do not move at all. Whereas the prescribed
smoothing performs excellent on this surface, the ring is a stable
limit surface of this flow.
Start Animation: prescribed
MCF, anisotropic MCF
Denoising
linear features The
octahedron is the
standard model for anisotropic smoothing algorithms. The images are
made to allow a comparison to results by others. Especially the
comparison shows the high quality of our results.
