Humans and many other animals have forward facing eyes with overlapping visual fields. Such animals appear to be able to extract "useful" information about the form and distance of objects in the external world quickly and efficiently, despite the unavoidable ambiguities in images and image pairs. However, it has proved difficult to develop methods for doing this artificially, mainly because of the problem of unambiguously matching up those points in the two images that correspond to the same point in the external world.
The results from two methods for the extraction of stereoscopic information will be described. Both of these methods are consistent with the known organisation of the visual system and the responses of its neurones. In one, features are extracted independently from the two images and these features are then matched with each other. In the second, the input from the two eyes is combined at different disparities and features are then extracted from this combined information.
Last modified: Thursday, 14-Oct-2004 14:23:21 NZDT
This page is maintained by Caroline Wills.