Image rectification for prism-based stereoscopic optical systems
| Autor: | A.O. Kuznetsov, Alexey V. Gorevoy, Alexander S. Machikhin |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Computer science
business.industry 3D reconstruction 020207 software engineering Stereoscopy 02 engineering and technology law.invention Optical axis Catadioptric system Rectification law Signal Processing Digital image processing 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Ray tracing (graphics) Computer vision Computer Vision and Pattern Recognition Artificial intelligence Image rectification business Software |
| Zdroj: | Computer Vision and Image Understanding. 182:30-37 |
| ISSN: | 1077-3142 |
| DOI: | 10.1016/j.cviu.2019.02.003 |
| Popis: | Image rectification is an important stage of digital image processing for three-dimensional (3D) coordinate measurements performed using stereoscopic technique. Rectification minimizes image distortions and radically reduces the calculation time necessary for stereo matching. Calibration of prism-based stereoscopic optical systems using the ray tracing camera model instead of a pinhole one leads to significant reduction of measurement errors, but this model has no single center of projection and does not allow applying conventional rectification techniques. This leads to the necessity of separate calibrations of the pinhole model for image rectification and the ray tracing model for 3D reconstruction. To overcome this contradiction, we propose the rectification technique, which uses the inverse ray tracing to find intersection points of the rays with the intermediate plane in the object space and then projects these points onto image planes of the virtual cameras. The distance to the rectification plane has significant impact on the misalignment of rectified images, so it was analyzed using computer simulation for various distances to the object. The technique has been integrated into the 3D model acquisition algorithm and verified by applying to the images of test and real objects obtained by two different miniature prism-based stereovision systems. For the typical working range 12–30 mm of endoscopic stereo imagers and the single rectification distance of 15 mm, it provides the rectification accuracy better than the accuracy provided by the pinhole rectification technique. Additionally, we have shown that the ray tracing camera model combined with the proposed rectification method can compensate the prism angular displacement around the optical axis, which hinders a correct matching of rectified images. The proposed technique is applicable to many non-central camera models used for catadioptric systems, depth-from-refraction systems and systems using lenses with significant pupil aberrations. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect