Sequential polarization imaging using multi-view fusion

Abstract

Multi-channel imaging like RGB is common today, however, a custom configuration of other modalities, such as wavelengths or polarization, is expensive and often unavailable. We propose a novel configurable multi-modal imaging system that solves this problem by using off-the-shelf components and proper fusion algorithms. The paper shows the application of polarization imaging for machine vision inspection. Sequential images acquired with different cameras and polarization angles are first registered and then fused together. With standard components, we demonstrate a superior image quality to existing specialized cameras while being significantly cheaper, available, and easily configurable.