3D Shape Representation
My ongoing work in 3D shape representation has been inspired by the
need to design more effective techniques for portraying the 3D spatial
relationships between multiple overlapping or enclosing surfaces.
Although transparency has the potential to be a useful device for
conveying information about the relative positions and orientations of
two or more superimposed layers --
in an ideal transparent rendering, one would be able to view the multiple
surfaces simultaneously and in their true context, and to rapidly intuit the
complex 3D geometry of the scene --
in practice, it is often difficult to adequately perceive the
full 3D shape of a smooth transparent shell or to accurately judge its
depth distance from underlying objects.
Over the past years, I have investigated a variety of techniques for
"artistically enhancing" transparent surfaces with sparse, opaque texture
markings that intend to describe the essential character of the surface shape
in a perceptually intuitive and minimally occluding way.
The images below illustrate some of my more recent results.
The
slides
from my talk at SIGGRAPH 97 provide a more complete, illustrated overview of this work.
click here for animated gif
(rotation sequence)- 2.0M (or on image for mpeg movie- 1.1M)
This pair of images contrasts a traditionally rendered transparent
surface with a rendering in which the transparent surface has been
enhanced by a carefully constructed pattern of thin opaque lines.
The intention of the lines is to both more explicitly define the 3D
location of the outer surface (via more reliable occlusion and vergence cues),
and to more clearly indicate its 3D shape.
The texturing technique illustrated in the image on the right is
described in my 1997 SIGGRAPH paper
"Illustrating Surface Shape in Volume Data via Principal Direction-Driven
3D Line Integral Convolution"
(2.4M pdf)
(HTML).
This image shows an example of some slightly earlier work, in which
I used the principal directions and principal curvatures at evenly
distributed points over an isosurface to define the orientation and length
of a polygonally-defined solid stroke texture. This texturing technique
and its perceptual motivation are described in
Interrante, Fuchs and Pizer.
"Conveying the 3D Shape of Smoothly Curving Transparent Surfaces via
Texture", IEEE Transactions on Visualization and Computer Grahics,
3(2):98-117
(3.3M pdf)
(HTML).
This reference also includes a description of the controlled observer
experiments that we conducted to evaluate the potential effectiveness of the
principal direction texturing approach.
click on the image above to see an mpeg rotation sequence- 1.5M
It's possible that principal direction stroke textures might have the
potential to be useful for depicting the shapes of complex objects for other
purposes, such as non-photorealistic rendering. There are a number of
obstacles that remain to be overcome before the techniques I've developed
could be viably used for such purposes, however. A foremost concern is
the problem of satisfactorily removing the kinds of directional inconsistencies
that are clearly visible across the broad upper areas on both sides of this
image.
Other major concerns include developing a more appropriate stroke character,
and improving the computational efficiency of both the principal direction
estimations and the generation and application of the texture across the
surface.
This image shows some of my earliest work in defining opaque texture
markings for illustrating surface shape.
Inspired by the ability of artists to define a figure with just a few strokes,
I sought to use the ridge and valley lines to define an orientation-invariant
3D "sketch" of the outer skin surface. The goal was to facilitate an immediate
understanding of the 3D location of the relevant soft tissue structures
indicated by the skin surface, without unduly obscuring the visibility of the
underlying tumor and dose surfaces.
This research is described in the paper
"Enhancing Transparent Skin Surfaces with Ridge and Valley Lines",
proc. Visualization '95, pp. 52-59
(728k pdf).
Last updated: 11/17/97
to my home page