Camera Calibration for Light Fields and Camera Arrays

Camera calibration, crucial for computer vision tasks, often relies on planar calibration targets to calibrate the camera parameters. This work explores a planar, fractal, self-identifying calibration pattern, which provides a high density of calibration points for a large range of magnification factors. An evaluation on ground truth data shows that the target provides very high accuracy over a wide range of conditions.


Fractal Calibration Target

The size of the calibration markers also play a key role in the accuracy of the localization. If too few calibration points are available per view, then calibration quality suffers due to over fitting. However, if the calibration markers become too small, it becomes impossible to properly detect them. As the optimal size depends on the camera magnification, it is not possible to create a pattern with a single optimal density.

Our solution is the adoption of a fractal scheme which operates on multiple scales of calibration points. This scheme allows to use features which are within some fixed bound of the maximum calibration point density for the given magnification, please see Figure. 1. In addition the target is self-identifying and quite robust under a range of imaging conditions.


Fig. 1: The recursive calibration target at several magnifications.


Fig. 2: Evaluation of the fractal marker pattern performance under radial distortion.

Fig. 3: Comparison of marker performance under Gaussian  noise and blur. The fractal pattern provides much higher precision but detoriorates under strong blur.

[1] A Fractal Calibration Pattern for Improved Camera Calibration Hendrik Siedelmann, Maximilian Diebold, Marcel Gutsche, Hamza Aziz-Ahmad, Bernd Jähne
Forum Bildverarbeitung, 2016[paper]