Recovering 3D depth information from two or more 2D intensity images is a long standing problem in the computer vision community. This paper presents a multi-baseline, coarse-to-fine stereo algorithm which utilizes any number of images (more than 2) and multiple image scales to recover 3D depth information. Several "view-selection strategies" are introduced for combining information across the multi-baseline and across scale space. The control strategies allow us to exploit, maximally, the benefits of large and small baselines and mask sizes while minimizing errors. Results of recovering 3D depth information from a human head are presented. The resulting depth maps are of good accuracy with a depth resolution of approximately 5mm. Comments University of Pennsylvania Department of Computer and Information Science Technical Report No. MSCIS-94-18. This technical report is available at ScholarlyCommons: https://repository.upenn.edu/cis_reports/529 View Selection Strategies for Mult i-View, Wide-Baseline Stereo MS-CIS-94-18 GRASP LAB 373 Hany Farid Sang Wook Lee Ruzena Bajcsy University of Pennsylvania School of Engineering and Applied Science Computer and Information Science Department Philadelphia, PA 19104-6389
[1]
Roger Y. Tsai,et al.
A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV cameras and lenses
,
1987,
IEEE J. Robotics Autom..
[2]
Jake K. Aggarwal,et al.
Structure from stereo-a review
,
1989,
IEEE Trans. Syst. Man Cybern..
[3]
Takeo Kanade,et al.
Virtual Space Teleconferencing Using a Sea of Cameras
,
1994
.
[4]
Paul R. Cohen,et al.
Camera Calibration with Distortion Models and Accuracy Evaluation
,
1992,
IEEE Trans. Pattern Anal. Mach. Intell..
[5]
Narendra Ahuja,et al.
Matching Two Perspective Views
,
1992,
IEEE Trans. Pattern Anal. Mach. Intell..
[6]
Takeo Kanade,et al.
A multiple-baseline stereo
,
1991,
Proceedings. 1991 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.