Our intention in this chapter has been to show the magnitude of the task carried out by the human vision system by getting down to the `nuts and bolts' of possible (although not always plausible) mechanisms. We have addressed the problem of designing a vision system for a mobile tourist guide and looked in detail at two ways of recognizing a familiar object: the London Underground sign. One way of recognizing a particular object in a TV image is to match a series of templates corresponding to idealized versions of the object, viewed from different angles and distances, against areas on the image. To compensate for variations in lighting, the image and the templates can be reduced to the form of edge maps: abrupt variations in light intensity that correspond approximately to the egdes of objects. Another approach is to extract a range of features, such as edges and regions, from the scene, and then to match these against models of possible objects in the scene, stored in the form of knowledge representation structures such as semantic nets.