Title: Image and video-stream encoding using near-optimal triangulations

Speaker:Georg Umlauf, University of Kaiserslautern

Time and Place: CSE 404, 4pm, Apr 6, 2007

Abstract:

In this talk I will present a method to speed up the optimization
approach to find a near-optimal triangulation to approximate an
image, using simulated annealing. Since the number of vertices of the
triangulation is small compared to the number of pixels, storing or
transmitting the triangulation saves a lot of space or bandwidth,
respectively. Furthermore, we define probability distributions that
speed up the convergence of the simulated annealing approach
significantly.

Since affine transformations of a triangulation can be computed very
efficiently, resizing or rotating the image can also be computed very
efficiently, which is useful for scaling from small displays (e.g.
handheld devices, PDAs, mobile phones) to large displays (e.g. CRTs,
power walls). Because our approximation is C0 continuous and a vector
based graphic format, resizing is possible without aliasing
artifacts.