In this paper, we propose an efficient stereo matching algorithm using morphological filtering and finger print on the scale space. We propose a morphological filter using a Gaussian structure element, which has lower computational complexity than conventional Gaussian filtering with similar performance. In stereo matching, we propose a coarse-to-fine feature-based method to minimize the effect of mismatching and noise by scale change. In the proposed stereo matching algorithm, we use the loci of zero-crossing points in the left and right images, as the robust matching features, and dynamic programming for feature correspondence. Computer simulation results with several test images show the effectiveness of the proposed feature-based stereo matching algorithm using the finger prints on the scale space.

Date Published: 21 April 1995
PDF: 8 pages
Proc. SPIE 2501, Visual Communications and Image Processing '95, (21 April 1995); doi: 10.1117/12.206714

Published in SPIE Proceedings Vol. 2501:
Visual Communications and Image Processing '95
Lance T. Wu, Editor(s)