Low-cost visible light spectral imaging

Abstract

Spectral imaging is an emerging technology for measuring spectral power distributions (SPDs) of electromagnetic radiation over a two-dimensional spatial domain. Within the visible light wavelength domain, spectral imaging measures light and colour with greater accuracy than digital photography. However, high cost limits its accessibility. Accordingly, a low-cost method was developed using commercially-available hardware – primarily a DSLR camera and a set of narrow bandpass filters. The quantity of filters was minimised to a total of seven, set by the dimensionality of SPDs, the spectral sensitivities of eyes and cameras, and commercial availability. Camera spectral sensitivity was measured using this same filter set, a colour chart, a spectrophotometer, and noon daylight modelled as CIE D65. The RAW photo format was used to access unprocessed sensor data. Independent SPD measurements from each colour channel were fused as a sensitivity-weighted average for efficient and continuous interpolation between colour channels with a high signal-to-noise ratio. Images were reconstructed from SPDs with standard observer functions. The method was demonstrated with a Canon 650D DSLR camera, a set of Thorlabs one-inch narrow bandpass filters, an X-Rite ColorChecker chart, and a Spectro 1 spectrophotometer. Accuracy was validated by quantitative comparison against ground truth SPD measurements and qualitative assessment of reconstructed images. The total filter cost was $715, plus $405 to measure camera sensitivity.