Abstract
The spatial correlation (SC) structure of a scene is highly informative about its content. For instance, if the SC structure is high, you are probably dealing with a scene with coherent figure-ground segmentation. Yet, when it is low, you are most likely looking at a very cluttered scene (Scholte et al., 2009). By fitting a Weibull distribution to the contrast histogram of an image, one can estimate two free parameters - beta and gamma, where beta closely resembles the contrast energy present in the image, while gamma is indicative of the SC structure of the scene. In recent work, we have shown that it is neurally feasible to estimate gamma from the pooled activity of lower-tier neuronal populations (Scholte et al., 2009). Here, we study to what extent the lateral occipital complex (LOC) of the visual cortex may be the locus of such computations. In a first experiment we established that BOLD-MRI activity in LOC evoked by natural scenes is well described by the gamma parameter. In a second experiment wwe observed that activity in LOC was sensitive to artificial stimuli that only differed in their gamma parameter. Retinotopic mapping showed that sensitivity to gamma is particularly pronounced in the posterior part of LOC (LO1/LO2). These data suggest that the posterior part of LO represents the spatial correlation of natural scenes, which provide a fast and efficient mechanism for analyzing the structure of the scene as a whole. We believe that this sheds new light on the proposed functions of the LOC in vision.
Meeting abstract presented at VSS 2013