Explains how to deal with the vast quantity of data required for computer vision by starting with a low spatial resolution (and therefore a small amount of data) and proceeding through successive refinements to the final verification of the results at the highest resolution available. Provides the g