Purchase this article with an account.
matteo valsecchi, Baptiste Caziot, Benjamin T. Backus, Karl R. Gegenfurtner; No role for early stereo in scene recognition. Journal of Vision 2012;12(9):447. doi: 10.1167/12.9.447.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Binocular disparity, like brightness and chromaticity, is invariably present in the scenes we encounter in our daily life. Yet, whether and how early stereo contributes to the visual processing of natural scenes is largely unknown. In a first experiment we presented observers with images of forests in 2D or 3D for 17 to 100 ms, followed by a (2D or 3D, respectively) mask. A 2AFC match to sample task followed. Disparity did not increase recognition performance for any presentation time. Similar results were found in a second study that used artificial grayscale scenes that consisted of several dozen randomly oriented cubes suspended in space. One explanation could be that disparity extraction is slow. To test this explanation we ran a third experiment using simple scenes created by Random Dot Stereograms with added luminance cues. In each scene, 8 noncentral locations contained patches defined by disparity and luminance contrast relative to the background. The task was again 2AFC match to sample, with distracters having a different configuration. Stereo coherence and contrast were titrated to provide 75% accuracy with 120 ms presentation. Reducing presentation time affected contrast and stereo pattern recognition similarly, and both patterns were still recognized above chance at 67 ms, suggesting that stereo extraction was not slow. Finally, we used a 2AFC in which one choice matched the contrast configuration and the "distracter" matched the disparity configuration. Participants gave some weight to stereo, but they chose the contrast pattern with higher probability than predicted based on relative visibility. In sum, our results show that stereo was available, but it was not used for fast recognition of the natural and artificial scenes we used. Instead, participants relied on features defined by other cues, such as luminance and color.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only