Structural stereo matching of Laplacian-of-Gaussian contour segments for 3D perceptionThe stereo correspondence problem is solved using Laplacian-of-Gaussian zero-crossing contours as a source of primitives for structural stereopsis, as opposed to traditional point-based algorithms. Up to 74 percent matching of candidate zero crossing points are being achieved on 240 x 246 images at small scales and large ranges of disparity, without coarse-to-fine tracking and without precise knowledge of the epipolar geometry. This approach should prove particularly useful in recovering the epipolar geometry automatically for stereo pairs for which it is unavailable a priori. Such situations occur in the extraction of terrain models from stereo aerial photographs.
Document ID
19900023979
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Boyer, K. L. (Ohio State Univ. Columbus, OH, United States)
Sotak, G. E., Jr. (Ohio State University Columbus, United States)
Date Acquired
August 14, 2013
Publication Date
January 1, 1989
Subject Category
Cybernetics
Meeting Information
Meeting: Optics, Illumination, and Imagae Sensing for Machine Vision III