This lecture presents research on a general framework for perceptual organization that was conducted mainly at the Institute for Robotics and Intelligent Systems of the University of Southern California. It is not written as a historical recount of the work, since the sequence of the presentation is not in chronological order. It aims at presenting an approach to a wide range of problems in computer vision and machine learning that is data-driven, local and requires a minimal number of assumptions. The tensor voting framework combines these properties and provides a unified perceptual organization methodology applicable in situations that may seem heterogeneous initially. We show how several problems can be posed as the organization of the inputs into salient perceptual structures, which are inferred via tensor voting. The work presented here extends the original tensor voting framework with the addition of boundary inference capabilities; a novel re-formulation of the framework applicable to high-dimensional spaces and the development of algorithms for computer vision and machine learning problems. We show complete analysis for some problems, while we briefly outline our approach for other applications and provide pointers to relevant sources.

Autorentext

Philippos Mordohai received his Diploma in Electrical and Computer Engineering from the Aristotle University of Thessaloniki, Greece, in 1998. He also received the MS and PhD degrees both in Electrical Engineering from the University of Southern California, Los Angeles, in 2000 and 2005, respectively. He is currently a postdoctoral research associate at the Department of Computer Science of the University of North Carolina in Chapel Hill. His doctoral dissertation work focused on the development of perceptual organization approaches for computer vision and machine learning problems. The topics he has worked on include feature inference in images, figure completion, binocular and multiple-view stereo, instance-based learning, dimensionality estimation, and function approximation. His current research is on the 3D reconstruction of urban environments from multiple video cameras mounted on a moving vehicle. Dr Mordohai is a member of the IEEE and the IEEE Computer Society, reviewer for the Transactions on Pattern Analysis and Machine Intelligence and the Transactions on Neural Networks. He served as chair of local organization for the Third International Symposium on 3D Data Processing, Visualization and Transmission that was held in Chapel Hill in 2006. Gerard Medioni received the Diplome d'Ingenieur Civil from the Ecole Nationale Superieure des Telecommunications, Paris, France, in 1977, and the MS and PhD degrees in Computer Science from the University of Southern California, Los Angeles, in 1980 and 1983, respectively. He has been with the University of Southern California (USC) in Los Angeles, since 1983, where he is currently a professor of Computer Science and Electrical Engineering, codirector of the Computer Vision Laboratory, and chairman of the Computer Science Department. He was a visiting scientist at INRIA Sophia Antipolis in 1993 and Chief Technical Officer of Geometrix, Inc. during his sabbatical leave in 2000. His research interests cover a broad spectrum of the computer vision field and he has studied techniques for edge detection, perceptual grouping, shape description, stereo analysis, range image understanding, image to map correspondence, object recognition, and image sequence analysis. He has published more than 100 papers in conference proceedings and journals. Dr Medioni is a Fellow of the IEEE and a Fellow of the IAPR. He has served on the program committees of many major vision conferences and was program chairman of the 1991 IEEE Computer Vision and Pattern Recognition Conference in Maui, program cochairman of the 1995 IEEE Symposium on Computer Vision held in Coral Gables, Florida, general cochair of the 1997 IEEE Computer Vision and Pattern Recognition Conference in Puerto Rico, program cochair of the 1998 International Conference on Pattern Recognition held in Brisbane, Australia, and general cochairman of the 2001 IEEE Computer Vision and Pattern Recognition Conference in Kauai. Professor Medioni is on the editorial board of the Pattern Recognition and Image Analysis journal and the International Journal of Computer Vision and one of the North American editors for the Image and Vision Computing journal.

Inhalt
Introduction.- Tensor Voting.- Stereo Vision from a Perceptual Organization Perspective.- Tensor Voting in ND.- Dimensionality Estimation, Manifold Learning and Function Approximation.- Boundary Inference.- Figure Completion.- Conclusions.