Direct evidence for occlusion in stereo and motion

Authors:
James J. Little;Walter E. Gillett
Affiliations:
Laboratory for Computational Vision, Department of Computer Science, University of British Columbia, Vancouver, British Columbia, Canada V6T 1W5;Antificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, USA
Venue:
Image and Vision Computing
Year:
1990

Citing 6
Cited 1

A Computational Approach to Edge Detection

IEEE Transactions on Pattern Analysis and Machine Intelligence
Algorithmic Techniques for Computer Vision on a Fine-Grained Parallel Machine

IEEE Transactions on Pattern Analysis and Machine Intelligence
The MIT vision machine

Artificial intelligence at MIT
Measurement of Visual Motion

Measurement of Visual Motion
Dynamic Occlusion Analysis in Optical Flow Fields

IEEE Transactions on Pattern Analysis and Machine Intelligence
Analysis of Accretion and Deletion at Boundaries in Dynamic Scenes

IEEE Transactions on Pattern Analysis and Machine Intelligence

Statistics of visual and partial depth data for mobile robot environment modeling

MICAI'06 Proceedings of the 5th Mexican international conference on Artificial Intelligence

Quantified Score

Hi-index	0.02

Visualization

Abstract

Discontinuities of surface properties are the most important locations in a scene; they are crucial for segmentation because they often coincide with object boundaries^1. Standard approaches to discontinuity detection decouple detection of disparity discontinuities from disparity computation. We have developed techniques for locating disparity discontinuities using information internal to the stereo algorithm of Drumheller and Poggio^2 rather than by post-processing the stereo data. The algorithm determines displacements by maximizing the sum, at overlapping small regions, of local comparisons. The detection methods are motivated by analysis of the geometry of matching and occlusion, and the fact that detection is not just a pointwise decision. Our methods can be used in combination to produce robust performance. This research is part of a project to build a Vision Machine^3 at MIT that integrates outputs from early vision modules. Our techniques have been extensively tested on real images.