Keywords
lightness, contrast gain control, assimilation, edge integration, phantom illusion
Abstract
In the ‘phantom’ illusion (Galmonte, Soranzo, Rudd, & Agostini, submitted), either an incremental or a decremental target, when surrounded by a luminance gradient, can to be made to appear as an increment or a decrement, depending on the gradient width. For wide gradients, incremental targets appear as increments and decremental targets appear as decrements. For narrow gradients, the reverse is true. Here, I model these phenomena with a two-stage neural lightness theory (Rudd, 2013, 2014) in which local steps in log luminance are first encoded by oriented spatial filters operating on a log-transformed version of the image; then the filter outputs are appropriately integrated along image paths directed towards the target. A contrast gain control mechanism adjusts each filter’s gain on the basis of the outputs of other nearby filters. The weighted contribution of each filter to the target lightness decays exponentially with distance, as do the strengths of the between-filter gain modulations. I simulate the lightnesses of incremental and decremental targets as a function of gradient width and show that the model reproduces the key properties of the phantom illusion, even when the gain applied to decremental luminance steps is 3 times larger than the gain applied to incremental luminance steps; and the inwardly- and outwardly-directed gain controls have opposite signs. These ancillary assumptions are motivated by previous data and theory (Rudd, 2010, 2013), but they are not required in order to reproduce the qualitative effects of wide and narrow gradients on incremental and decremental target appearance.
Start Date
15-5-2015 9:50 AM
End Date
15-5-2015 10:15 AM
Session Number
05
Session Title
Early Vision
Included in
Cognition and Perception Commons, Computational Neuroscience Commons, Systems Neuroscience Commons
‘Edge’ integration explains contrast and assimilation in a gradient lightness illusion
In the ‘phantom’ illusion (Galmonte, Soranzo, Rudd, & Agostini, submitted), either an incremental or a decremental target, when surrounded by a luminance gradient, can to be made to appear as an increment or a decrement, depending on the gradient width. For wide gradients, incremental targets appear as increments and decremental targets appear as decrements. For narrow gradients, the reverse is true. Here, I model these phenomena with a two-stage neural lightness theory (Rudd, 2013, 2014) in which local steps in log luminance are first encoded by oriented spatial filters operating on a log-transformed version of the image; then the filter outputs are appropriately integrated along image paths directed towards the target. A contrast gain control mechanism adjusts each filter’s gain on the basis of the outputs of other nearby filters. The weighted contribution of each filter to the target lightness decays exponentially with distance, as do the strengths of the between-filter gain modulations. I simulate the lightnesses of incremental and decremental targets as a function of gradient width and show that the model reproduces the key properties of the phantom illusion, even when the gain applied to decremental luminance steps is 3 times larger than the gain applied to incremental luminance steps; and the inwardly- and outwardly-directed gain controls have opposite signs. These ancillary assumptions are motivated by previous data and theory (Rudd, 2010, 2013), but they are not required in order to reproduce the qualitative effects of wide and narrow gradients on incremental and decremental target appearance.