Abstract
How do we ascertain the shapes of unfamiliar objects? Why do we so seldom confuse shadows with real things? How do we "factor out" shadows when looking at scenes? How are we able to see the world as essentially the same whether it is a bright sunny day, an overcast day, or a night with only streetlights for illumination? In the terms of this paper, how can we recognize the identity of Figs. 2.1 and 2.2? Do we use learning and knowledge to interpret what we see, or do we somehow automatically see the world as stable and independent 'of lighting? What portions of scenes can we understand from local features alone, and what configurations require the use of global hypotheses?