Representations of Video and 3D Data for Manipulation and Recognition Harpreet S. Sawhney Vision Technologies Laboratory Sarnoff Corp., Princeton, NJ Abstract Video cameras are no longer being used only in their traditional role of providing "viewable pixels", but are rapidly becoming sources of intelligent information about the world. More recently 3D cameras/sensors are being developed to directly provide 3D measurements of objects and scenes. Appearance and geometry of objects and scenes, and the temporal dynamics of objects are the key information bearing sources for visual information. This talk will highlight the representation of dynamic scenes in terms of 2D and 3D layered representations. Applications ranging from tracking to enhanced video visualization for tele-presence and video object fingerprinting will be used to illustrate the work. I will also present highlights of recent work on classification and recognition of 3D objects from view-based 3D sensor data. I will describe a structural coarse-to-fine indexing and matching approach for scaleable 3D object recognition. Bio Dr. Harpreet S. Sawhney - Technical Manager, leads the Vision Technologies group at the Sarnoff Corporation. Dr. Sawhney received his Ph.D. in Computer Science in Feb. 1992 from the University of Massachusetts, Amherst, focusing on Computer Vision. His areas of interest are 3D Modeling, Vision & Graphics Synthesis, Video Enhancement, Video Indexing, Data Mining and Compact Video Representations. Since 1995, he has led government and commercial programs in Immersive Telepresence, Image based 3D Modeling, Video Object Fingerprinting, Video Mosaicing, Geo-registration, 2D and 3D Video Manipulation, and Object Recognition. Dr. Sawhney was one of the key technical contributors towards the founding of two Sarnoff spinoffs, VideoBrush Inc., and Lifeclips Inc. Between 1997 and 2002, he was awarded the Sarnoff Technical Achievement Awards five times for his contributions in Video Mosaicing, Video Enhancement, 3D Vision and Immersive Telepresence. Between 1992 and 1995 he led video annotation and indexing research at the IBM Almaden Research Center in San Jose, CA. Dr. Sawhney has served on the Program Committees of numerous Computer Vision and Pattern Recognition conferences. He has published over 50 papers and holds 11 patents.