Relational Matching for Stereopsis
Loading...
Date
1991-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Ohio State University. Division of Geodetic Science
Abstract
Extensive research has been carried out to understand stereopsis in order to solve the correspondence problem. The motivation for investigating the correspondence problem in digital photogrammetry is the fact that it is still far from being solved. The most prominent problems are relief distortions, occlusions, discontinuities in the surface (breaklines) and nonlinear radiometric differences. Applications such as large scale urban areas or close range scenes with a large depth range relative to the base/height ratio often pose unsurmountable problems to existing matching methods. This work is concerned with solving the correspondence problem by using a relational description of features rather than the features themselves. The objectives are to match signals derived from intensity values along epipolar lines, using as few constraints and assumptions as possible, and to determine as many conjugate points as possible. Moreover, the matching scheme should produce reliable results even if large numbers of geometric and radiometric distortions are present. In this research, relational matching in the form of tree matching is used. To demonstrate the flexibility of the proposed matching scheme, two input signals were used. The first input consists of gray levels along epipolar lines used to match gray levels, and the second input involves convolution values used to match (indirectly) zero-crossings. Special attention was given to match scenes with large amounts of occlusions, foreshortening, and break lines and to use as few constraints as possible. The results confirm the theoretical expectations that relational matching successfully copes with large geometric distortions without introducing special constraints or tuning the algorithm.