Abstract
The lightness and the depth perceptions are two fundamental properties of vision. Importantly, it is possible that the perception of images are established by dynamic interactions of the processes underlying the lightness and depth perception in the visual system. However, how these two computational processes interact is unknown. A neurocomputational model called DISC model (differentiation integration for surface construction) was developed to compute "border-ownership" signals that indicate the figural side at a border (Kogo et al., 2010, Psych. Rev.). The model has been improved further by implementing the detection of long-range consistency of surface properties and showed robust responses. In this new model, the interaction between the depth and the lightness computations can be made in the form of "differentiated signals", i.e. the border-ownership (difference of depth) and the contrast (difference of achromatic color), and the depth and lightness values can be computed by the 2D integration of these signals. In the well-known lightness illusion, "White's illusion", the perceived lightness of the target surfaces differs from what is expected by contrast enhancement suggested by the simultaneous contrast illusion. Involvement of the depth order perception in the lightness computation to explain this illusion has been suggested (Ripamonti & Gerbino, 2001, Perception). Hence, the illusion is the ideal case to test the approach described above. We tested the model not only with the classic White's illusion but also with the "inverted White's illusion" (Spehar et al., 2002, Perception). Because of the complexity of the figure and the gradual reversal of the figure-ground organization in the perception of the image, the image was split into two vertical halves. With this configuration, the computed lightness was shifted to the directions matching the psychophysical data. We report the details of the algorithm and the responses of the model.
Meeting abstract presented at VSS 2016