Nonlinear models of colour vision

The world about us is vibrant and rich with colour - a woodland walk on a sunny autumn day, for example, presents us with seemingly infinite shades of stunning autumnal colours. Our daytime visual experience, and colour experience in particular, is derived from just three different kinds of light-sensitive cells in the retina (cones) that respond differently to different wavelengths of visible light, i.e., to the different bands in the rainbow.

From a scientific standpoint. then, colour vision should be simple and straightforward, but it has puzzled some of the greatest scientific minds for centuries. In part, this is because the problem of colour vision itself is much more complicated than it first appears. The composition of the light that reaches our eyes depends on the material properties of the objects upon which the light falls - a leaf absorbs significant amounts of both shortwave (blue appearing) and longwave (red appearing) light and so appears green, while roses are red and violets are blue because they reflect relatively more long and shortwave light, respectively.

But it also depends on the composition of the light source itself - the white light of the sun on a cloudless day, for example, or the orange glow of an incandescent lightbulb. One of the striking features of colour vision is that appearance varies relatively little with these large changes in the illumination, despite the consequent changes in the cone responses.

To accomplish this the visual system must incorporate information across time and space into the process of extracting and categorizing colour from cone responses.

We propose to use experimental techniques to measure the complex ways in which the colour appearance of lights depend on their context in time or space. On a computer monitor or projection screen we will show human observers flickering coloured lights that vary from each other in precisely controlled ways and ask the observer some simple questions about the appearance, such as 'which of two lights looks redder?' or ask them to adjust the colour of one light in order to match another.

We will then analyse their patterns of responses and how they change in relation to changes in the visual stimulus in order to build simple models of visual processing that link the input (what was shown on screen) to the output (what the observer perceived). These models will in part be based on results from the wealth of literature about how individual cells in the eyes or brains of animals respond to different kinds of light stimulation to better understand how colour is sequentially processed in humans in which those kinds of invasive procedures are not possible.