VASC Seminar Announcement
            =========================

Date: Wednesday, 9/8/99
Time: 3:30-4:30
Place: Smith Hall 2nd Floor Common Area

Speaker:  Steve Seitz
          CMU Robotics Institute
          http://www.cs.cmu.edu/~seitz


Title: Plenoptic Scene Reconstruction

Abstract:

The manner in which a scene reflects light provides implicit information about 
surface geometry and reflectance.   By studying the pattern of light radiated 
from a scene, we may therefore hope to estimate its three-dimensional 
structure.  Generally, we can capture only a subset of this light.  Suppose, 
however, that we could acquire photographs of a scene from every viewing 
position and orientation.  What could be inferred about the structure of the 
scene from these images and via which algorithms?  

The set of all possible images of a scene, a space called the plenoptic 
function, is of fundamental importance to the field of computer vision, 
because it provides a complete description of scene appearance.  While 
capturing the complete plenoptic function is not generally possible, we can 
sample it by taking numerous photographs.  In this talk, I will describe how 
best to sample the plenoptic function for the purpose of 3D scene 
reconstruction, and how to compute 3D geometry from such a sample using a 
new volumetric approach called Space Carving.  This is joint work with Kyros 
Kutulakos, Harry Shum, and Adam Kalai.