Abstract
Purpose>> Several studies have shown that disparity information provided by the non-vertical spatial frequency (SF) components in a stimulus exert a substantial influence on the perception of stereoscopic depth. Here, we tested how non-horizontal disparity information contributes to the sensitivity of the human stereovision system by comparing stereothresholds for orientationally broad- and narrow-band stimuli with equal contrast in their vertical SF components. Methods>> Stereothresholds were measured in 2 subjects for SF band-pass (2–4 cpd) and orientation low-pass filtered random-dot stimuli (mean luminance = 3 cd/sq. m), using the method of constant-stimuli. Broad-band stimuli included SF components at all orientations from vertical (0 odeg) to the cut-off orientation of 89 odeg. Narrow-band stimuli with a cut-off orientation of 15 odeg were tested for comparison. All SF components that did not belong to the selected spatio-orientation band were removed. To maintain orientation balance, a corresponding band of negative orientations was present in each stimulus. The RMS contrast was 0.21 cd/sq. m for broad-band stimuli and 0.15 cd/sq. for narrow-band comparison stimuli. These values produce equal contrast of the broad-band and narrow-band stimuli within the 15 odeg orientation band. Results>> Stereothresholds for the broad-band stimuli are lower by a factor of approximately 2.5, compared to those obtained with the near-vertical narrow-band stimuli. Similar results were obtained when the RMS contrast of the broad-band stimuli was 0.31 cd/sq. m. Conclusions>> The results provide clear evidence that optimal stereo-sensitivity depends on disparities in SF components outside the ±15 odeg band. We conclude that disparity signals from vertically and non-vertically tuned neural mechanisms are pooled prior to computation of horizontal image disparity and that this pooling substantially enhances the signal to noise ratio of the stereovision system.
Supported by R01 EY05068 and R01 MH 49892