On the Calibration of Focused Plenoptic Cameras

Plenoptic cameras provide a robust way to capture 3D information with a single shot. This is accomplished by encoding the direction of the incoming rays with a microlens array (MLA) in front of the camera sensor. In the focused plenoptic camera, a MLA acts like multiple small cameras that capture the virtual scene on the focus plane of a main lens from slightly different angles, which enables algorithmic depth reconstruction. This virtual depth is measured on the camera side, and independent of the main lens used. The connection between actual lateral distances and virtual depth, however, does depend on the main lens parameters, and needs to be carefully calibrated. In this paper, we propose an approach to calibrate focused plenoptic cameras, which allows a metric analysis of a given scene. To achieve this, we minimize an energy model based upon the thin lens equation. The model allows to estimate intrinsic and extrinsic parameters and corrects for radial lateral as well as radial depth distortion.