Abstract
There are many ways in which humans can judge depth and shape: stereo, shading, motion, etc. Presumably the outputs of the different depth mechanisms are used by the brain to create a true 3D representation. We modified the classification image technique and were able to visualise the observer's 3D representation by means of the “classification object”. The method was the same as for a classification image, except that we used the z-coordinate instead of luminance at each x,y location in the computation. The observers' task was to discriminate a human face (shown by 2500 random points on its surface) from a version that was stretched along the z-axis. On each trial, the observer judged whether the face or the z-stretched version rotated about a vertical axis. The resulting classification object has a peak coinciding with the tip of the nose, and troughs at the corner of the eye and at the juncture of the lips. In a second experiment, we constructed a classification object for a sinewave depth grating. The troughs and peaks of the classification object coincide with those of the grating. We conclude that humans use a true 3D template when making judgements about 3D object structure.
Supported by a Discovery Grant from NSERC