The concept of smooth generic camera calibration for optical metrology

Modern digital cameras with sophisticated optics enable accurate shape measurements via a plethora of metrological techniques. In order to take advantage of the data delivered by high-resolution cameras, one needs a sufficiently flexible model to describe the imaging geometry and an adequate calibration procedure to determine the model parameters and estimate their uncertainties. In this report, we suggest the concept of a smooth generic camera model that has several advantages compared to the state-of-the-art: the description is resolution-independent and takes advantage of the natural smoothness of the projection geometry; the global and the local symmetries inherent to the view ray-based parametrization are regularized in an explicitly unbiased manner. The calibration procedure uses dense datasets that can be obtained with the technique of active screens and consistently accounts for the uncertainty in the registration data that can be extracted during pre-processing. The concept is illustrated with synthetic examples based on physically-accurate rendering.