A short paper.

ZombieBoard can be operated via a web browser. Or, in order to make whiteboard scanning convenient and casual in an office or conference room setting, a ZombieBoard can be operated via a Diagrammatic User Interface, whereby the user issues scan commands by drawing on the whiteboard itself.

The ZombieBoard application integrates four interesting technical components:

• Image Mosaicing: To obtain a high-resolution image of the whiteboard, the program must stitch together slightly overlapping snapshots of a sparsely marked 2d surface. I use a novel feature-based mosaicing method.
  
  
• Real-time Computer Vision:The Diagrammatic User Interface component must segregate people moving around in front of the whiteboard from marks on the whiteboard itself. This component was developed in conjunction with summer intern, Dietmar Aust.
  
  
• Line Drawing Analysis: My core research techniques of token grouping in scale-space blackboard are used to perform the line drawing analysis needed to recognize buttons, check marks, and other elements of user commands drawn as diagrams on the whiteboard.
  
  
• Distributed Object Architecture: Image mosaicing is a compute-intensive process, best run on a powerful compute-server. By contrast, camera control and frame grabbing are best performed by scan servers local to each installation. In order to support a scalable system with many whiteboard scanners controllable via Diagrammatic User Interfaces and web browsers, we designed a distributed object oriented client/server architecture, implemented in ILU. Summer intern Bikram Bakshi built this component of the system.

Example image (published in
N. Streitz, J. Siegel, V. Hartkopf, S. Konomi (Eds.),
Proceedings of CoBuild'99. Second International Workshop on
Cooperative Buildings, Pittsburgh. LNCS 1670. Springer: Heidelberg.)