Image Mosaicing and a Diagrammatic User Interface
for an Office Whiteboard Scanner

Eric Saund
Xerox Palo Alto Research Center


This paper presents two computer vision components underlying a novel camera-based whiteboard scanner now in routine use at our research center. First, a fully-automatic feature-based mosaicing algorithm enables high-resolution whiteboard imaging using a video resolution pan/tilt camera, requiring minimal frame overlap even in the presence of sparse curvilinear data. Second, real-time activity detection coupled with grouping-based line drawing analysis enable a Diagrammatic User Interface whereby commands are issued to the system by drawing on the whiteboard itself. This application represents a step toward reducing the barrier between the physical and electronic workspaces.

Postscript (7.6M compressed to 980K)

PDF (907K)