Biomedical Engineering Reference
In-Depth Information
Most of the surface representation schemes found in literature have adopted
some form of shape parameterization especially for the purpose of object
recognition. However, free-form surfaces, in general (e.g., CT/MRI skin con-
tours), may not have simple volumetric shapes that can be expressed in terms
of parametric primitives. Some representation schemes for free-form surfaces
found in literature include the
“splash”
representation proposed by Stein and
Medioni [27] in which the surface curvature along the intersection of the sur-
face and a sphere centered at the point of interest is calculated for different
sphere diameters and a signature curve is obtained for this point. Also the work
of Chua and Jarvis [28], who proposed the
“point signature”
representation
which describes the local underlying surface structure in the neighborhood of
a point. This is obtained by plotting the distance profile of a circle of points
to a plane defined by that circle of points. Dorai and Jain [29] proposed an-
other representation called
“COSMOS”
for free-form objects in which an ob-
ject is described in terms of maximal surface patches of constant shape index
from which properties such as surface area, curvedness and connectivity are
built into the representation. Johnson and Hebert [30] recently introduced a
new representation scheme called the
“spin image”
. This image represents
the histogram of the surface points relative to a certain oriented point. This
image is generated for each oriented point on the surface and matching be-
tween two surfaces is done by matching the spin images of the points in the
two surfaces. Yamany and Farag [26] introduced another technique based on
surface signatures
. Surface signatures are 2D images formed by coding the
3D curvature information seen from a local point. These images are invariant
to most transformation. In what follows are some details for some of these
algorithms.
1.5.1.1
The “Splash” Surface Registration
Geometric indexing have been one of the most used surface indexing techniques
because it used the geometrical relationships between invariant features. How-
ever, another form of indexing that uses local shape information has become
more popular. As it is based on structural information local to the neighborhood
of a point, this indexing method is called “Structural indexing”.
Circle of points was first used to describe the underlying surface structure
around a given point. This can be done by decomposing the local surface patch
