This is a non-exclusive list of projects I believe would be
interesting to explore, take a decent amount of effort to
implement, and might even lead to some interesting research
questions. Basically the idea is to pick stuff from recent state
of the art research. Here goes:
- Digital Cockroach Farm: Build and animate a
hexapod. Then have a whole gang of them with group behaviors
such as flocking, hunting, etc. Make it all run on the
workbench as an autonomous animation. There is relevant
literature in the A-life community as well as in recent
Siggraph papers on flocking behavior.
- Mesh Compression: Describe large meshes efficiently
Topological Surgery. Involves juicy graph
algorithms. Currently the best algorithm for
non-progressive mesh encoding. Can you improve on
this? Potential for original research.
- View Interpolation through Epipolar Morphing: Take
separate images and use view
morphing to create realistic inbetween images and/or use the
ideas to do
plenoptic image editing. High neatness factor. Don't
know where this is going to lead...
- Very high resolution terrain roaming: Be the first on your
block to roam fluidly over a 7k by 10k sample (10m
dataset of the Los Angeles basin. Requires mesh
simplification in the regular setting. Putting this on the
workbench would be a lot of fun. Chance to play with some of the
neatest datasets on this planet (no kidding) and make a
substantial contribution to application sciences. Come see
me about papers in this direction.
- Large Image Mosaics: Implement algorithms which can
stitch together many images into
large scale image mosaics. Think about enhancing such a
viewer with depth information to do a better job at rerendering.
- Weighted Mesh Simplification: Apply weighted,
constrained meshing algorithms to images as a representation
primitive. Neat and funky. Will lead to original
research. Talk to me about details. If you know your
numerical analysis, we'll solve some operators. Very high
- Subdivision, Wavelets and the Lifting Scheme: Create
an interactive (JAVA?) toolbox for the lifting scheme and
use it explore non-uniform variational
subdivision and associated wavelet transforms for an
interactive editing system. Potential for net.fame.