Abstract
There is convincing evidence that motion parallax resulting from head movements and stereopsis use similar, if not the same, neural architecture. Given the geometric similarities it is tempting to assume that other types of motion-defined form are processed in the same manner. However, we have previously shown that 3D form from motion in a kinetic depth display (KDE) relies on additional shape cues (ie. texture gradient), while stereoscopic form perception remains unaffected by such information. These results suggest that there may be important differences in the processing of motion and stereoscopically-defined form. Here we test this hypothesis by comparing the resilience of form perception in the presence of depth noise. The amount of rotation required to reliably detect a depth-defined form in random-element stereograms and kinematograms was assessed for four subjects. Thresholds were measured for three mean levels of depth noise (1.3, 5.1 and 7.6 min) at each of three standard deviations (0, 1.3 and 3.82 min). Our results show that the effect of depth noise on thresholds for form recognition was markedly different in the motion and stereoscopic conditions. That is, motion thresholds increased directly with the mean and standard deviation of the noise. Stereoscopic thresholds were not affected by either of these variables. In addition, the overall variability of the threshold measurements was greater in the motion conditions than in the stereoscopic conditions. It is clear from these data that the stereoscopic system is more resilient to external noise. This result, taken with our previous results showing that the KDE is also less robust to reductions in element density and exposure duration argue in favour of distinct early processing of these two types of form information. We postulate that form recognition is aided by a surface interpolation signal in the stereoscopic system that is not available when 3D form is defined by motion.
K. J. MacKenzie is supported by an NSERC Postgraduate Scholarship. This work was conducted with funding support from NSERC to L. M. Wilcox