Vision Sciences Society 2006
Volume 6, Number 6, Abstract 655, Page 655a doi:10.1167/6.6.655 http://journalofvision.org/6/6/655/ ISSN 1534-7362
Computations underlying fine and coarse stereopsis
Takahiro Doi Laboratory for Cognitive Neuroscience, Graduate School of Frontier Biosciences, Osaka University, Japan
Seiji Tanabe Laboratory for Cognitive Neuroscience, Graduate School of Frontier Biosciences, Osaka University, Japan
Ichiro Fujita Laboratory for Cognitive Neuroscience, Graduate School of Frontier Biosciences, Osaka University, Japan
In the early stage of stereo processing, neurons encode the disparity energy of stereo images. These neurons appear to encode the depth of images without a global-match solution to the stereo correspondence problem. In higher processing stages, responses to false matches are attenuated in the ventral pathway, while the disparity-energy signal is passed on to the dorsal pathway. To understand how these two types of depth information contribute to stereopsis, we examined human psychophysical performance in depth discrimination at various binocular disparity levels. Subjects were required to discriminate the depth of a random-dot stereogram that comprises two groups of dots: one with the same contrast between the two eyes, and the other with the opposite contrast. Altering the percentage of dots belonging to each group allowed us to test the computation underlying stereo performance at each disparity level. In trial blocks in which the disparity was large, subjects perceived no depth when the percentage of the two groups was equal, and reversed depth when the opposite-contrast group dominated. In blocks in which the disparity was small, subjects perceived correct depth when the percentage of the two groups was equal, and no depth when the opposite-contrast group dominated. We conclude that the two depth representations contribute to stereopsis in a manner that depends upon the disparity scale demanded by the task. The disparity energy-like representation is used for coarse scale discrimination, while the false-match-attenuated representation is employed for fine depth discrimination.
Supported by grants from the MEXT (17022025) and the Takeda Science Foundation