Distortion in Multiple View Geometry

For modern consumer cameras often approximate calibration data is available, making applications such as 3D reconstruction or photo registration easier as compared to the pure uncalibrated setting. However, images obtained online sometimes have no calibration information, and never have information about the distortion model. We present methods to estimate focal length, distortion center and one distortion parameter in various settings. We address calibrated-uncalibrated image pairs with known distortion center (for one image intrinsic parameters are assumed to be known, whereas the second view has unknown focal length and distortion), calibrated-uncalibrated image pairs with unknown distortion center, totally uncalibrated image pairs with unknown distortion centers (both images have unknown intrinsics and both have distortion) and image pairs from the same camera with unknown calibration and distortion model.

Publications

• One-sided Radial Fundamental Matrix Estimation (BMVC 2012)
• Unknown Radial Distortion Centers in Multiple View Geometry Problems (ACCV 2012)
• Radial Distortion Self-Calibration (CVPR 2013)

Software

• One-sided Radial-Fundamental Matrix with known distortion center demo (Matlab)
• Radial-Fundamental Matrices with unknown distortion center demo (Matlab)

Contact

• José Henrique Brito
• Kevin Köser