Ashish Sabharwal
|
This artifact has 35 hand-made key frames converted to a 15 second movie
using Catmull-Rom interpolation. Multiresolution editing was used to handle
varying levels of details (prominent in the beginning and near the end). A
flight like effect is obtained by varying the number of inbetween-frames
between pairs of adjacent key frames.
|
|
Vassili Sukharev
|
Here we see another movie created with the subdivision surface
animation system jointly developed with Ashish Sabharwal. Shot by
Vassili, this movie stars Ashish early in the morning at work.
|
|
Gretta Bartels and Alex Mohr
|
Our artifact illustrates what we've learned this quarter by combining
all of the projects into one rendering. One of our artifacts from the
Impressionist project and another of our artifacts from the first
ray-tracing project are each texture mapped onto squares. An object
from the subdivision surfaces project is rendered in the background.
The contributions of our final project, texture mapping of diffuse
color and transparency (Gretta) and distribution ray-tracing for
reflection and area lighting (Alex) are demonstrated throughout the
artifact. The third square contains a colored transparency texture
map which allows only colored portions of the gray background to peek
through. The fuzzy shadows of all of the objects illustrate the area
lighting. The fuzzy reflection of the subdivision surface shows the
distribution ray tracing.
|
|
Ken Yasuhara
|
This flower model was automatically generated from a small
L-system grammar and consists of cylinders and flattened,
stretched spheres.
|
|
Matthew Cary
|
Direct Subdivision Surface Editing
|
|
Brian Tjaden
|
View Dependent Adaptive Level of Detail for Subdivision Surfaces
Using a view-dependent measure for each face of an object, I adaptively
subdivide that face. "Back" faces which the viewer cannot see are not
subdivided at all. "Silhouette" faces are highly subdivided (up to the
user specified maximum level). "Front" faces which more directly face the
viewer are subdivided moderately.
The main obstacle which needed to be overcome was subdividing adjacent
faces at different levels. A little clever local subdivision resolved
this problem; however, this can cause some minor flaws along border
faces (faces whose adjacent faces are subdivided at a different level)
if care is not taken.
As the artifacts show, image quality does not
suffer with adaptive view-dependent subdivision, but there is significant
reduction in subdivision computation. The adaptive view-dependent
approach requires between 1/2 the number of faces (when the user specified
maximum level of subdivision is 1) and 1/10 the number of faces (when the
user specified maximum level of subdivision is about 4) compared to the
number of faces subdivided in the non-adaptive approach.
|
|
Steve Capell
|
These animations were generated using physical simulation on objects
represented as subdivision lattices. Lattices provide internal
structure that does not exist for surface representations, thus giving
a volumetric feel to the animations. Two kinds of deformation are
supported, one in which an object prefers its original shape, like a
rubber eraser, and one in which an object simply prefers its original
volume, like blueberry jelly
|
|
Yung-Yu Chuang
|
Based on Taubin's signal processing view on surface fairing, I
integrate the non-shrinking fairing algorithm into subdivision surface
modeler. It can handle hierarchical and smooth deformation constraints
as well.
|
|
Chris Prince and David Ely
|
This project extends the subdivision surface renderer. We use view-dependent
information to reduce the number of polygons that have to be drawn, while
maintaining image quality.
The information we consider includes (a) the object's silhouette, (b) the
object's curvature, (c) the location of specular highlights, (d) the final
screen-size of the polygons, and (e) whether polygons are front- or back-facing.
|
|
Jonathan Nowitz
|
I extended Impressionist to do multiresolution automatic
impressionizing. The set of brushes used is parameterizable, as is
the amount of Guassian blurring to apply before each coat, the
tolerable error between the blurred image and the canvas before a new
stroke will be applied, and the size of the "buckets" that are
examined to determine whether or not to put a stroke in the bucket.
Based on the SIGGRAPH 98 paper, but since Impressionist already has a
variety of different strokes and additional features, its a little
more flexible (though probably not more realistic in its artifacts).
|
|
Mathieu Blanchette
|
The images were obtained using texture mapping and bump mapping. The two
images illustrate different depth of field, with elements out of focus
being blurred.
|
|
Valentin Razmov
|
Briefly named "dunes" (although you'll see it as Valentin.bmp) this artifact
shows the potential for building non-selfsimilar looking surfaces, which in our
world translates to randomly shaped objects like mountains, clouds, fire, water,
Brownian motion, climate modeling, etc.
The idea behind this is fractal subdivision - a simple scheme where the for
each two points in the mesh, the generated midpoint is protruded noisily with
some vector in the 3rd dimension (the "height"). This gives the sense of
jaggedness (of peaks), if the protrusion vector is large, or more smooth hills
in the case of a smaller protrusion vector. Most often, that protrusion vector
is scaled based on a 2D distance (after a projection onto the x-y plane) between
the two even points, which originally spawned the odd vertex. That gives a sense
of balance in the shapes. Also, for more realistic mountain effects, the
protrusion can depend on the "height" - in this case areas of higher altitude
will be more jagged, just like peaks normally are, while areas of valley terrain
will look smoother.
|
|
Jeffrey Hightower and Daniel Lloyd
|
We implemented a generalized set of preprocessor instructions to
allow for motion over a set of frames based on a subdivision curve of
predefined control points. Next, we added support in our ray tracer
for general texture mapping of surfaces. Lastly, we implemented a
physical energy transfer modeller to simulate the kinetic collisions
that would occur on a 9-ball break. The modeller also handles friction
of collisions and motion and warps the ball rotation matrix to simulate
rolling along the table.
|