Blur and Disparity Provide Complementary Distance Information for Human Vision

Abstract

Disparity is generally considered the most precise cue to depth, while blur is considered a coarse, qualitative cue. Depth from disparity and depth from blur have similar underlying geometries: one is based on triangulation between images collected by different eyes and the other is based on triangulation between images collected through different parts of the pupil. Thus, from a geometric standpoint, they provide complementary distance information. Physiologically, the two cues have very different sensitivities. Disparity thresholds, expressed as just-noticeable differences in depth, are low near fixation, but increase rapidly away from fixation. In contrast, blur thresholds are relatively large and do not vary significantly with position relative to fixation. Thus, one might expect disparity to determine depth discrimination near fixation and blur to determine discrimination away from fixation. We tested this expectation in a psychophysical experiment. Observers were presented a reference and test stimulus on each trial. The two stimuli were either both in front of fixation or both behind fixation. After each trial, observers indicated which stimulus appeared more distant. We used a novel volumetric display (Love et al., 2009) to present stimuli that contained 1) only disparity information (Gaussian dot viewed binocularly), 2) only blur information (a disk with 1/f noise viewed monocularly), or 3) both disparity and blur information (1/f disk viewed binocularly). As expected, thresholds were lower in the disparity-only condition than in the blur-only condition and the two-cue thresholds were similar to the disparity-only thresholds when the reference and test were near fixation. The situation reversed, however, behind fixation where blur thresholds were lower than disparity thresholds and two-cue thresholds were similar to blur-only thresholds. Thus, disparity and blur are complementary sources of information with disparity providing the best depth information near fixation and blur providing the best information away from fixation.