Robust stereo vision

Saravajit Sahay Sinha; Saied Moezzi; Brian G. Schunck

doi:10.1117/12.25365

1 March 1991 Robust stereo vision

Saravajit Sahay Sinha, Saied Moezzi, Brian G. Schunck

Proceedings Volume 1385, Optics, Illumination, and Image Sensing for Machine Vision V; (1991) https://doi.org/10.1117/12.25365
Event: Advances in Intelligent Robotics Systems, 1990, Boston, MA, United States

Abstract

This paper presents an algorithm for visual surface reconstruction which makes use of a robust local reconstruction scheme to produce a dense disparity map from a multiresolution feature-based stereo matching algorithm. Robustness implies that the algorithm can reject large amounts of outliers in the disparities which are caused by mismatches in the correspondence process while simultaneously preserving discontinuities in the depth. Our robust algorithm uses a standard multiresolution stereo algorithm in conjunction with a moving least median of squares (MLMS) algorithm to fit local planar patches to the disparity functional at each level of the multiresolution pyramid. The MLMS algorithm finds the best fit by minimizing the median of the error between the fit and the data. By applying the MLMS algorithm at each stage of the pyramid we not only create a denser grid at each level but also " nip in the bud" any errors which occur at a coarse level before they are propagated to finer levels of the multiresolution process. Experimental results are presented on real and synthetic data.

Citation Download Citation

Saravajit Sahay Sinha, Saied Moezzi, and Brian G. Schunck "Robust stereo vision", Proc. SPIE 1385, Optics, Illumination, and Image Sensing for Machine Vision V, (1 March 1991); https://doi.org/10.1117/12.25365

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available