The interpretation of structure from motion is examined from a computational point of view. The question addressed is how the 3-D structure and motion of objects can be inferred from the 2-D transformations of their projected images when no 3-D information is conveyed by the individual projections.
