Texture difference cues in figure-ground separation

Parsing an image into figure and ground is a critical element of visual processing, and can be driven by textural cues. These textural cues can be of many different types, including differences in the distribution of luminance levels, differences in orientation, and differences in higher-order image statistics. Recently, we showed that for differences in orientation content, there is an asymmetry between figure and ground: thresholds depend not only on the magnitude of the orientation difference, but also on whether the oriented component is present in the figure vs. the ground. Since, for these stimuli, the figure-ground distinction is only possible after orientation is analyzed, this asymmetry implies functional recursion. Moreover, this figure-ground asymmetry was not present for differences in un-oriented spatial frequency content, i.e., granularity. Here, we extend this analysis to other kinds of textural differences. In a 2-AFC task, subjects identified a target image, consisting of five randomly-positioned circular disks rendered by one synthetic texture superimposed on a background rendered by a second synthetic texture. The non-target image consisted of a uniform texture whose image statistics matched the spatial average of the target’s. All images consisted of 64 x 64 arrays of 10-min black and white checks. The circular disks of the target collectively covered 25% of its area. Texture construction enabled separate control of the fraction of black and white checks, oriented and non-oriented spatial frequency content, and higher-order (third- and fourth-order) local image statistics; 16 such combinations were examined. Figure-ground asymmetries were small when target disks were defined by luminance differences alone or higher-order image statistics alone. When orientation differences were combined with differences in higher-order statistics, figure-ground asymmetries were enhanced. This enhancement supports the notion that figure-ground analysis utilizes functional recursion from visual areas that process higher-order images statistics to orientation processing in V1.