A bird's-eye view of color patterns: Inferring spectral reflectance from multi-illuminant imaging

Color vision differs dramatically across animal species. For example, many birds have tetrachromatic vision, with one of the four cones being sensitive to ultraviolet light. To study coloration patterns in nature, such as butterfly wing patterns, it is important to understand how they are perceived by different animals. Color can be objectively quantified in terms of spectral reflectance, which can then be projected onto the cone sensitivities of a given animal. Past studies have used either spectrometric methods with limited spatial resolution, or digital photography with low spectral resolution. We have developed a method to infer spectral reflectance from digital camera images, enabling characterization of color patterns with high spectral and spatial resolution. Our method involves imaging an object under different light sources and applying an optimization algorithm to reconstruct the reflectance spectrum at every pixel. We have tested the method on both simulated data and real images of butterfly specimens, which allowed us to infer how they would look to potential predators, such as birds.