All wavelengths are equal but some are more equal than others.

Graham Finlayson

(University of East Anglia, Norwich)

The famous Newton Prism experiment appears to indicate that the colour we see depends on the (dominant) wavelength of light. This is in fact not true. The same physical stimulus may appear blue, green or red depending on the context in which it is seen. Yet, the idea of wavelength surely matters in colour vision and it appears that some wavelengths are more crucial than others. Specifically, the prime wavelengths of light around 450, 540 and 610 Nanometres (the visible spectrum runs from about 400 to 700 Nanometres) appear most important.

In the talk I show the following results. First, that a monitor that has its red, green and blue primaries anchored at the prime wavelengths is optimally energy efficient. Second, that such a monitor induces the largest gamut of colours. Third, that a camera designed to drive a prime light monitor has attractive colour balance properties (specifically it easy to model and remove the effect of viewing illuminant). Finally, by applying the same reasoning to our own visual systems we come up with a physical basis for perceptual response (this idea might be seen as controversial).

So, why is this interesting to a computer science audience? Well, these ideas are now well established in computer vision (see papers on color constancy and photometric invariants). Moreover, the prime wavelength concept is being incorporated in camera standards. In recent work we have generalised the prime light idea to multispectral data in computer graphics . Indeed, the prime light idea is the key to fast multispectral rendering (which in turn is the key to photo realism). Other possible applications for prime lights will also be discussed.
Thursday 28th November 2002, 14:00
Robert Recorde Room
Department of Computer Science