Binocular correspondence is the study of how the monocular images are correctly fused in normal binocular vision, creating a sense of depth. In natural scences, there is usually considerable ambiguity about the binocular matching of local elements within each monocular half-image. For the past three years, we have used simple repetitive patterns (wallpaper stereograms) to explore the rules which control human stereo matching. Stereo matching of a long horizontal row of points can be described as follows: coarse features, such as edges or gaps in the row, are matched first. A plane interpolated between the positions in depth assigned to these coarse features then guides the matching of the intervening points. What is the mechanism of this interpolated matching? The proposed research will address the following questions: 1) Does the stereo system use the unambiguous matches (the edges) detected by coarse spatial filters to weight the numerous potential matches detected by fine spatial filters selectively? 2) What role does the relative contrast within different spatial bandwidths play in the relative weighting of the matches detected by coarse and fine spatial filters? 3) Are brightness and orientation also important dimensions in stereo matching? 4) If stereo matching is based on local cross-correlation of the monocular half-images, what is the size and disparity range of this operation? 5) Can we find evidence for signals in many depth planes using an adaptation paradigm? A final series of experiments will examine the relationship between incremental judgments of disparity (delta d/d) and incremental judgments of monocular distance (delta s/s) to determine if contemporary models of hyperacuity can be extended to the disparity domain as well.
|Effective start/end date||31/12/89 → 31/12/07|
- National Eye Institute
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.