Stereo Matching in the Presence of Sub-Pixel Calibration Errors

Stereo matching commonly requires rectified images that are computed from calibrated cameras. Since all under- lying parametric camera models are only approximations, calibration and rectification will never be perfect. Addi- tionally, it is very hard to keep the calibration perfectly sta- ble in application scenarios with large temperature changes and vibrations. We show that even small calibration er- rors of a quarter of a pixel are severely amplified on cer- tain structures. We discuss a robotics and a driver assis- tance example where sub-pixel calibration errors cause se- vere problems. We propose a filter solution based on signal theory that removes critical structures and makes stereo al- gorithms less sensitive to calibration errors. Our approach does not aim to correct decalibration, but rather to avoid amplifications and mismatches. Experiments on ten stereo pairs with ground truth and simulated decalibrations as well as images from robotics and driver assist...

• Type: Conference
• Year: 2009
• Where: CVPR
• Authors: Heiko Hirschmüller, Stefan K. Gehrig
• Keywords: Computer Vision, CVPR 2009, Stereo Matching, Sub-Pixel Calibration