Figure 2.
Retinal adaptation of input scene followed by pre-attentive boundary and surface processing in the 3D ARTSCAN model. Light adaptation at the model's outer segment of the photoreceptors and spatial contrast adaptation at the inner segments of photoreceptors are implemented as in the aFILM model (Grossberg and Hong, 2006) (Equations 1–8). The outputs from the inner segment of the photoreceptors input to the model LGN. These inputs are contrast-normalized by single opponent networks of ON and OFF cells via on-center off-surround and off-center on-surround interactions, respectively, among cells that obey membrane equation, or shunting, dynamics (Equations 9–14) and then by double-opponent networks (Equations 15, 16). LGN double-opponent outputs are used to compute orientationally- and contrast-selective simple cells that are selective to four different orientations (Equations 17–20). Simple cell outputs are pooled across all four orientations to yield complex cells (Equation 21). Complex cells, in turn, input to monocular left (L) and right (R) eye retinotopic boundaries.