Image-based 3D surface reconstruction by self-consistent fusion of shading
and shadow features
Christian Wöhler
DaimlerChrysler Research and Technology
Ulm
Mittwoch, 14.07.2004, 16 Uhr c.t., D6-135
In this talk a framework for image-based three-dimensional surface
reconstruction
by self-consistent fusion of shading and shadow features is presented.
Relying
on the analysis of at least two pixel-synchronous greyscale images of the
scene
acquired under very different illumination conditions, this framework
combines
a shadow analysis of the first image of the scene, allowing for a
determination
of large-scale altitude differences on the surface at high accuracy, with a
variational shape from shading scheme applied to the second image (and
eventually
to further images), estimating the surface gradients and altitude profile.
In a
first step, the result of shadow analysis is used for selecting a solution
of the
variational shape from shading scheme which is consistent with the average
altitude
difference derived by shadow analysis. In a second step, the detailed
shadow
structure is taken into account. An error term that aims at adjusting the
altitude
differences extracted from the reconstructed surface profile to those
derived from
shadow analysis is incorporated into the error function to be minimized by
the
variational shape from shading scheme. The second reconstruction step is
initialized
with the result of the first step. Moreover, this approach is extended to
the
analysis of two or more shadows observed under different illumination
conditions,
which yields an initialization of the shape from shading algorithm.
In contrast to traditional shape from shading or photometric stereo
approaches,
the presented framework shows the advantage that it (a) alleviates the
ambiguity
of the solution of the shape from shading problem, (b) does not require a
very
accurate knowledge about the reflectance function of the surface to be
reconstructed,
and (c) permits a coplanar configuration of the illumination vectors.
The described framework is applied to the astrogeological task of
three-dimensional
reconstruction of regions on the lunar surface using ground-based CCD
images, as
well as to three-dimensional reconstruction of metallic surfaces.
Zur Startseite