Figure 4. The linear complex-cell feedforward network model.

(A) The key components of the three-layer complex-cell model are illustrated. Layer 1 (purple): the receptive field (RF) of each V1 cell (c) is characterized by four two-dimensional Gabor functions ($G_c^i$, i = 1:4, each having a distinct spatial phase 0, 180, 90 and 270 degrees, respectively). The response of each model V1 cell’s RF to an input image, I, is quantified as the half-wave rectified dot product between the image and the corresponding Gabor function $\left(R_c^i=G_c^i\cdot I\right)$. Layer 2 (cyan): a complex V1 cell’s response to the input image is modeled by summing responses of the four RFs modeled in layer 1. Layer 3 (khaki): the response of a V2 pixel/cell to the input image, R_p, is modeled as a weighted sum of all inputs from V1. ω_cp represents the weight for a V1–V2 cell-pixel pair (B) Spatial distribution of various parameters employed in constructing the RFs of two distinct V2 cells/pixels in case MK373-CTB647. These cells/pixels exhibit POs of 45° (TOP) and 90° (BOTTOM). In these plots, each projecting V1 cell is depicted as an ellipse, aligning with the orientation and aspect ratio of its associated Gabor function. The colors used here denote the weights within the linear model.