Purchase this article with an account.
H. Steven Scholte, Sennay Gebreab, Arnold Smeulders, Victor Lamme; Visual gist of natural scenes derived from image statistics parameters. Journal of Vision 2009;9(8):1039. doi: https://doi.org/10.1167/9.8.1039.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Natural images are highly structured in their spatial configuration. In the past it has been shown that the contrast distribution of natural images is almost always adequately described by a Weibull type distribution (Geuseboek & Smeulders, 2003) in which 2 free parameters are fitted. We have recently shown that these parameters explain up to 50% of the variance in the early ERP and these parameters correlate 0.84 and 0.93 with the modeled output of X and Y cells of the LGN (Scholte et al., submitted). Here we will present BOLD-MRI data that show that beta and gamma also explain single trial activity in the occipital and temporal cortex and the parietal cortex respectively. Also, the beta and gamma parameters seem to order the natural images along the dimensions of the number of objects that are present in the scene and depth organization of the scene. We will test this hypothesis by estimating beta and gamma for artificial stimuli with a pre-determined number of objects and depth organization, and by evaluating brain responses to such stimuli. Our results indicate that the summary statistics of the Weibull distribution (beta and gamma) may be used by the brain to efficiently and very rapidly extract information about the visual gist of natural scenes.
This PDF is available to Subscribers Only