Abstract
In the Benary cross illusion two gray triangles are perceived on a black cross against a white background. Interestingly, the placement along the cross and the orientation of the gray triangles crucially determines their perceived brightness, even though both the triangles, and the local contrasts at their borders are identical. This phenomenon has heretofore resisted mechanistic explanation. Local, isotropic filtering models, which have enjoyed considerable success in predicting other brightness illusions such as White's effect or the Adelson checkerboard, fail to correctly predict psychophysical data obtained with this stimulus in different orientations. Here we present a modified version of Blakeslee and McCourt's ODOG model that succeeds in explaining the Benary cross illusion and modifications of it. A key insight embodied in our modified model is that oriented filter outputs should be weighted according to psychophysically determined orientation sensitivity (contrary to the original ODOG formulation, which weighted oriented responses according to the space-averaged RMS contrast level). Not only does our model explain the psychophysical data from the Benary cross illusion, but it also retains the explanatory power of the original Blakeslee and McCourt model, thus providing a valuable formulation for predicting how we perceive brightness in the world.
Meeting abstract presented at VSS 2016