Scholte H S, Ghebreab S, 2011, "Summary statistics of model neural responses at different stages of the visual hierarchy" Perception 40 ECVP Abstract Supplement, page 208
Summary statistics of model neural responses at different stages of the visual hierarchy
H S Scholte, S Ghebreab
A fascinating feat of humans is their ability to rapidly categorize the gist of a scene. A reasonable assumption is that this (partly) relates to the statistical structure of natural images. In the past it has been show that the distributions of contrast values in natural images generally follow a Weibull distribution, with beta and gamma as free parameters. Beta describes feature energy while Gamma describes spatial coherency. Here we show that this also applies to distributions of edge orientation. Furthermore we show that when going from the retinal ganglion cells to V1 these parameters seem to structure visual images in an increasing meaningful way. If we plot the feature energy and spatial coherence derived from the retinal ganglion cells, we observe a differentiation between images with a coherent figure-ground segmentation from cluttered images while beta and gamma are correlated. In the LGN we observe that beta and gamma are de-correlated and gamma indicates the presence of a coherent figure-ground segmentation irrespective of the feature energy. In V1 we observe a more complex grouping of scenes. To validate this model we covaried the EEG responses of subjects viewing natural images with the beta and gamma values of those images. All models have a peak of their maximum explained variance around 109 ms with values of 0.59 for the retinal ganglion model, 0.81 for the LGN model and 0.69 for the LGN model. Remarkably, the V1 model explains up to 31% of the variance in the ERP signal after 200 ms while the explained variance of the other models rapidly drops after 120 ms. Together we believe this data shows that the feed-forward anatomy of the visual system provides the brain with meaningful summary statistics of the perceived scenes.
These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2011 Abstract Supplement (complete) size: 2206 Kb
[Publisher's note: The abstracts in this year's ECVP supplement have been published with virtually no copy editing by Pion, thus the standards of grammar and style may not match those of regular Perception articles.]