Detection of surfaces in structure from motion.

Abstract

In 3 experiments the authors examined the effects of surface shape on the detection of surfaces in structure from motion displays. In experiment 1 they found the number of points necessary to distinguish simple quadratic surfaces from points randomly placed in a volume to be low, in the range of 4-6 points. In experiment 2 they found that a hyperbolic shape is harder to detect than an arch or parabolic shape and that more curved shapes are easier to detect than less curved shapes. In experiment 3 they found that detection of sinusoidal shapes increases with increasing amplitude and decreasing frequency, confirming previous results with motion parallax displays. This pattern of results is consistent with a model in which triangular surface patches are fit through adjacent triples of points and displays with smaller mean orientation change between neighboring triangles are judged to be surface displays.