Sensorimotor Interaction in Stereopsis
Stereopsis is the reconstruction of depth from slight differences in the two retinal images. The first step in stereopsis is stereo-matching, the locating of corresponding visual features in the two eyes' retinal images. Stereo-matching is a demanding task, but it could be simplified by exploiting a geometric constraint: matching features are constrained to retinal bands called epipolar lines.
The epipolar lines slide on the retina when the eyes move, so eye position information is necessary if the epipolar constraint is to be used. Using random- dot stereograms it is shown that human stereo-matching does not shift retinal search regions along with the moving epipolar lines, but searches large retina- fixed two-dimensional retinal patches instead.
The size of these patches depends on the pattern of eye movements. In simulations of eye movement effects on the search zone size it is demonstrated that human eye movement follows a movement pattern that reduces the size of these search zones. This movement pattern is a compromise between motor advantages and the demands of stereopsis.
It is also shown that the visual system could use the epipolar constraint with retina-fixed search regions by using visual information alone to compute the location of the epipolar lines. This process is called visual bootstrapping. Psychophysical data are presented showing that the visual system does use visual bootstrapping.
The fusional horopter is the region of space where objects are seen single. A mathematical extension of the classical theoretical point-horopter is presented that allows the computation of the point-horopter as a full two- dimensional surface.
This dissertation shows that it is necessary to study the interaction of sensory and motor systems in order to fully understand the purpose of both. It uses the interaction of stereopsis and the oculomotor system as an example.