|
A yellow banana looks yellow to us whether we see it in the tungsten light of the kitchen or sunlight outdoors. This fact tells us that our brains perform the remarkable feat of colour constancy. Under different illuminations, the banana reflects very different light spectra to our eyes, but our brains correct for this difference and we see instead the unchanging material properties of the banana. That is, we see that the banana tends to reflect more light in the 'yellowish' region of the spectrum than other objects do, regardless of the type of light shining on them. It is still something of a mystery as to how our brains achieve colour constancy, but we do know that there is more than one mechanism involved. The simplest mechanism is re-adjustment (adaptation) of the light receptors in the eye in response to changes in the overall illumination -- if the amount of 'red' light increases overall, the receptors reduce their sensitivity to 'red'. But more complex mechanisms may also be involved that depend on, for example, object recognition and memory. If we recognise a particular object as a banana, we may remember its typical yellow colour, and use that memory colour to adjust its present colour, as well as to correct the colours of other objects under the same illumination. One reason for the mystery is that it is difficult to measure colour constancy in the laboratory. Traditionally, scientists have studied colour perception using artificial flat patches that are entirely uniform in colour and brightness. These are especially easy to control and display using computers. But they are very unlike real coloured objects in the natural world and they cannot tell us much about the role of object recognition and memory. We argue that if we want to understand colour perception in the real world, we need to use real objects with natural surfaces, and natural tasks. Therefore, we have developed a novel technique that allows us to change the surface colours of real, solid objects. We use a computer-controlled light projector, real objects, and hidden mirrors. With this setup, we have already found that the yellow of a yellow banana is indeed different and more stable than the otherwise same yellow of an unfamiliar yellow disk. But until now, we have only used uniformly coloured objects. We now want to make our objects even more realistic by adding texture to their surfaces. Look closely at a banana and you will see that it is not uniform yellow, but many different colours, ranging from bright yellow to dark brown, from tiny to large spots (depending on how ripe it is). We will improve our setup so that we can apply this sort of non-uniform natural colour texture to a range of objects. We will then be able to carry out new experiments to discover whether a naturally textured banana is even more stable in colour. We predict that the colour of a banana-textured banana will be better remembered, and more stable under illumination changes, than the colour of banana-textured disk, or an untextured banana. To do this work, we will devise new computer programs for making natural textures, projecting them onto 3D objects, and collecting and analysing responses from human subjects. The experiments require more than 100 subjects over three years to spend in total more than 1000 hours gazing at our objects and making rapid replies as to whether they are the same objects they saw before, or whether their colours have changed, and a variety of other questions. Of course we will also test many objects other than bananas, and therefore we will try to discover in general how our brains combine the perception of colour, texture, and shape when we recognise objects. On a practical level, our research will also help computer scientists to make more realistic computer graphics, and it will help manufacturers of products such as laundry detergents and shampoo understand how their consumers really perceive faded blue jeans and dyed hair.
|