Purchase this article with an account.
Arash Yazdanbakhsh, Margaret Livingstone; Neural dynamics of surface processing in V1. Journal of Vision 2007;7(9):327. doi: 10.1167/7.9.327.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Visual information processing requires the grouping of different parts of an image into discrete objects; one key step in this is surface segmentation. Surface segmentation and figure-ground segregation require integration of information across large distances in the visual field, yet they occur early in the perceptual process. In order to reconcile the requirement for integration across the visual field with the small receptive fields of cells in early visual areas, it has been suggested that a process of filling-in occurs early in the visual system, by which edge information propagates cell-to-cell from cells representing the borders of a surface towards cells representing its center. The existence or absence of filling-in in primary visual cortex, however, is controversial (Zipser, Lamme et al. 1996; Lee, Mumford et al. 1998; Friedman, Zhou et al. 1999; Lamme, Rodriguez-Rodriguez et al. 1999; Sasaki and Watanabe 2004; Heinen, Jolij et al. 2005; Cornelissen, Wade et al. 2006). Here by constructing a neural population response to edges and their enclosed surfaces using a novel reverse correlation method, we measured the spatio-temporal evolution of responses across a surface and its background. For small figures, responses to the inside versus the outside differed, suggesting a surface representation resembling filling-in. Increasing the surface size, however, eliminated the difference. Convolving the local receptive-field properties of cells with the different surfaces showed that what appeared to be a filling-in phenomenon could arise from the first order spatio-temporal kernel of V1 cells rather than from mechanisms dedicated to diffusion processes or filling-in. We suggest a biologically plausible model based on the receptive field spatio-temporal kernel of V1 cells. Predicting surface and edge responses based on such a kernel indicates that the surface representation and contour processing arise from the same mechanism.
This PDF is available to Subscribers Only