High-precision polarization measurements and analysis for machine vision applications

Polarization is a source of information that is steadily attracting attention in the field of computer vision due to its ability to tap into information not readily available in standard colour or greyscale cameras. Unfortunately, most existing data capture methods tend to suffer from either poor signal-to-noise ratio or long capture times. Further, most existing literature relies on making heavy assumptions about the polarizing properties of surfaces, which limits their application. This paper aims to optimise image capture conditions for polarization data in order to maximise the signal-to-noise ratio. Using the discovered optimal settings, a variety of images of different scenes are captured illustrating a range of reflectance properties typically overlooked previously. Such phenomena include inter-reflections, combined specular-diffuse reflection and surface conductance. The output from the paper is a set of key requirements and considerations necessary to further advance the field of polarization vision.