We acquire a great deal of information about the world about us by perceiving the colour of objects. We can manage without colour, but certain tasks can become rather difficult telling ripe from unripe fruits, or finding green tennis balls lost in grass, for example. Yet colour information is seldom used by the machine vision community, although, given the present state of image understanding technology, it is undesirable to waste cues as to the nature of the world observed. The colour signal received by an imaging device does not itself tell us much about the world. We need to compute from this signal the surface reflectances of the objects that were imaged. To do so, we need to be able to find changes in the colour signal. This requires an understanding of what a noteworthy difference in colour is. We discuss metrics on the colour space, colour constancy, and some models of colour vision. We describe a working colour edge finder, based on an opponent coding scheme, which uses Fleck's edge finder to generate a colour edge map.