Goal of the thesis is to develop an estimation technique for diffuse surface reflectance via an augmented reality user interface on mobile devices. Specifically, the goal is to not only obtain colour coordinates for the area of interest, but to make an educated guess which real-world spectral reflectance in a large database of paint samples best matches the surface in question. There are three cornerstones to the technique: first, the device camera will be calibrated at each measurement location via a colour chart, to properly estimate ambient illumination. Second, a database of measured spectral reflectances will be used as a ground truth to compare colour estimates against: and third, if ambient light levels permit, the light source of the mobile device (and possibly also other external light sources of known spectral emissivity) will be used to obtain a measurement under a second illumination. The third step serves to further disambiguate similar spectral responses, and to raise the accuracy of the results. The planned system will be integrated into an existing mobile library application that permits browsing of the existing spectral database, and that already allows finding of similar colours under varying illuminants.
