Evaluation of computational radiometric and spectral sensor calibration techniques

Abstract

Radiometric and spectral calibration are essential for enabling the use of digital sensors for measurement purposes. Traditional optical calibration techniques require expensive equipment such as specialized light sources, monochromators, tunable filters, calibrated photo-diodes, etc. The trade-offs between computational and physics-based characterization schemes are, however, not well understood. In this paper we perform an analysis of existing computational calibration schemes and elucidate their weak points. We highlight the limitations by comparing against ground truth measurements performed in an optical characterization laboratory (EMVA 1288 standard). Based on our analysis, we present accurate and affordable methods for the radiometric and spectral calibration of a camera.