Imagine you have arrived at one of London's railway stations and now wish to continue your journey by the Underground. You are confronted by the scene shown in figure 8.1.
As a seasoned traveller you know that a sure way of finding the entrance to the Underground station is to look for the London Underground logo (figure 8.2), and so you scan the scene before your eyes until the sign is spotted and then begin to move towards it, avoiding obstacles in your path, until the entrance is reached.
How does your visual system make this possible? Certainly, it seems to require very little effort, and it is tempting to believe that whatever is going on inside your head must be quite different from the mechanisms proposed in the previous chapters of this book. This chapter shows that, on the contrary, internal representations (chapter 3), search strategies (chapter 4), and structural descriptions (chapter 5) all play a central role in recent computational models of visual perception. For example, we show that the same depth-first search strategy used to plan a route through the London Underground system (chapter 4) can be used to recognize the London Underground sign in a street scene.