engleski

A feedback model of figure-ground assignment

Figure-ground organization is an important step in visual processing which separate structured input to which processing efforts should be devoted (figure) from less structured background. Gestalt psychologists identified several variables which influence figure-ground assignment including size, contrast, convexity and symmetry (Palmer, 1999). For instance, smaller surfaces tend to be perceived as figures as well as surfaces with larger contrast. Recently, two new factors affecting figure-ground organization were revealed. Those are lower region and top-bottom polarity. Lower region refers to the tendency to assign figure to the surface in the lower part of the visual field (Vecera et al., 2002). On the other hand, top-bottom polarity refers to the tendency to assign figure to the surface with wide base and narrow top (Hullemann & Humphreys, 2004). We propose architecture for visual processing which can account for classical and new principles of figural assignment. The model is based on the feedback interaction between what and where processing streams (Bullier, 2001). In particular, it is suggested that where processing stream computes saliency of the presented surfaces and feeds back this information into construction of surface representation. The model consists of three network layers. First layer computes oriented luminance discontinuities (edges) simulating the properties of simple and complex cells in the primary visual cortex. Second layer simulates the computation of saliency in the parietal cortex. This is achieved by convolution of summed output from the first layer with Gaussian kernels with wide spread. Finally, signals from the first and the second layers are combined in a recurrent network which provides surface representation (Domijan, 2004). Recurrent network implements object-based lateral inhibition and enables only one connected group of cells to remain active. This group of cells corresponds to the surface of perceived figure. Computer simulations showed that the model correctly assigns figural status to the surface with smaller size, larger contrast, bottom location and heavy bottom part. Key component to achieve this is a difference in activity across second layer. Stronger activity in the second layer is observed in the area with greater density of edge responses from the first layer making it more salient. When second layer activity is fed to the recurrent network it will bring competitive advantage to such area. Therefore, the model explains classical and new cues of figural assignment in a unified way using the same set of parameters. It provides biologically plausible implementation of the idea that top-down signals related to the attention have an active role in figure-ground assignment.

Vision; Figure-ground perception; Neural model

nije evidentirano