In their day-to-day lives, people naturally understand and operate in a three dimensional world. Curiously, though, they only sense 2D projections of it. The seemingly effortless act of inferring 3D from 2D observations is the result of complex mechanisms that are still quite far from being resolved. For many years, this task has been considered the primary role of visual processing. Pioneers in the fields of artificial intelligence and computer vision set out to recover a 3D representation of visible scenes which could then be used to recognize objects and reason about the world.
However, the general problem of recovering 3D from 2D imagery and the many steps involved 1 require a significant understanding of how the mind works, from issues of learning to intelligent behavior. Thus, the field is plagued by several of the same hurdles that have occupied AI researchers for many years. A tractable and more theoretically well-posed problem is the specific computation of 3D geometry from 2D geometry or Structure-from-Motion (SfM).