Abstract

Several lines of evidence indicate that stereoscopic information and motion information strongly interact to specify object shape and layout in depth. For example, adaptation to an unambiguous stereoscopic object temporarily stabilizes perception of an otherwise ambiguous kinetic (KD) display. In fact, stereoscopic adaptation can even create the perception of structure and depth in a display consisting entirely of dynamic random noise. Conversely, perceived depth in a KD display can bias depth segregation in a stereoscopic display. These and related findings have inspired a simple neural model that incorporates KD and disparity processing within a single network. The model comprises two levels: a layer of monocular directionally selective motion detectors that provide input to a second layer of disparity selective and direction selective binocular mechanisms. The network of facilitatory and inhibitory connections between binocular mechanisms promotes segregation of activity into separate neural pools specifying global motion in different depth planes. In the absence of disparity information, activity fluctuates in a manner mimicking the fluctuations in perception experienced when viewing a KD animation sequence. Unresolved questions include the role of head movements in modifying activity within the network.

© 1991 Optical Society of America