Using digital cameras to investigate animal colouration: estimating sensor sensitivity functions

Abstract

Spectrophotometers allow the objective measurement of colour and as a result are rapidly becoming a key piece of equipment in the study of animal colouration; however, they also have some major limitations. For example, they can only record point samples, making it difficult to reconstruct topographical information, and they generally require subjects to be inanimate during measurement. Recently, the use of digital cameras has been explored as an alternative to spectrophotometry. In particular, this allows whole scenes to be captured and objectively converted to animal colour space, providing spatial (and potentially temporal) data that would be unobtainable using spectrophotometry; however, mapping between camera and animal colour spaces requires knowledge of the spectral sensitivity functions of the camera’s sensors. This information is rarely available, and making direct measures of sensor sensitivity can be prohibitively expensive, technically demanding and time-consuming. As a result, various methods have been developed in the engineering and computing sciences that allow sensor sensitivity functions to be estimated using only readily collected data on the camera’s response to a limited number of colour patches of known surface reflectance. Here, I describe the practical application of one such method and demonstrate how it allows the recovery of sensor sensitivities (including in the ultraviolet) with a high enough degree of accuracy to reconstruct whole images in terms of the quantal catches of an animal’s photoreceptors, with calculated values that closely match those determined from spectrophotometric measurements. I discuss the potential for this method to advance our understanding of animal colouration.