Human stereovision without localized image features

Abstract

Many theories of human stereovision are based on feature matching and the related correspondence problem. In this paper, we present psychophysical experiments indicating that localized image features such as Laplacian zerocrossings, intensity extrema, or centroids are not necessary for binocular depth perception. Smooth one-dimensional intensity profiles were combined into stereograms with mirror-symmetric half-images such that these localized image features were either absent or did not carry stereo information. In a discrimination task, subjects were asked to distinguish between stereograms differing only by an exchange of these half-images (ortho- vs. pseudoscopic stereograms). In a depth ordering task, subjects had to judge which of the two versions appeared in front. Subjects are able to solve both tasks even in the absence of the mentioned image features. The performance is compared to various possible stereo mechanisms. We conclude that localized image features and the correspondences between them are not necessary to perceive stereoscopic depth. One mechanism accounting for our data is correlation or mean square difference.

A psychophysical and computational analysis of intensity-based stereo

“…In recent experiments with smoothly shaded gray wedges, we have shown that stereo disparities in continuous graylevel images can be detected even though salient image features were missing and, consequently, point disparities could not be defined (Arndt et al 1995). In a systematic test of a number of potential candidates for feature matching, we showed that neither luminance edges nor luminance extrema nor the overall centroid of a gray wedge are necessary to perceive stereoscopic depth.…”

“…Stimuli lacking Laplacian zero-and level-crossings as well as other features have been discussed at length in Arndt et al (1995). The simplest one is a parabolic intensity profile, the second derivative of which is constant.…”

“…As an underlying mechanism, Arndt et al (1995) suggested the minimization of the mean square difference of the image profiles by varying the horizontal offset between the images. This mechanism, which is equivalent to maximizing area correlation, is well in line with psychophysical experiments and with computational models of disparitytuned units (see above).…”

“…In other words, contrast and disparity should be reciprocally related to each other at threshold; e.g., if contrast is halved, disparity must be doubled to restore visibility. In experiment 1, this prediction is tested using the parabolic gray wedge introduced by Arndt et al (1995).…”

“…15 of Arndt et al 1995). D * is an average of the true disparities weighted with I 2 x , i.e., a measure of local image contrast in the x-direction.…”