Depth from monocular images

Abstract

Nakayama and Shimojo coined the term “Da Vinci stereopsis” for the impression of depth created by parts of a display seen by only one eye. In displays used to illustrate Da Vinci stereopsis the magnitude of depth is unspecified because the viewer has no information about the size of the occluded object. We have designed displays that remove this ambiguity. In one display a square has the same colour as a transparent background. For one eye, the square just fills a slit in the background so that its lateral edges are not visible. For the other eye, the square is displaced relative to the slit. For this eye, both lateral edges of the square are visible because the background is depicted as transparent. We refer to this depth cue as ‘monocular transparency’. We measured the depth produced by our display relative to that produced by an adjacent disparity depth probe. For most subjects, the perceived depth of the transparency display matched that of the probe. Thus, monocular transparency can create good quantitative depth in the absence of binocular disparity. We have also produced quantitative depth in displays resembling Panum's limiting case and show that Da Vinci stereopsis and Panum's limiting case are essentially the same. Our effects cannot be due to vergence-induced disparity because they occur when both images are symmetrical and because depth in both directions occurs at the same time.