Fig 1.
Examples of brightness induction.
(a) Brightness contrast: the grey squares have the same luminance, but the one over a black background appears lighter, and the other appears darker. (b) Brightness assimilation: all the grey bars have the same luminance, but the ones over a black background appear darker, while the ones over white appear lighter.
Fig 2.
Lateral inhibition process in the retina and the formation of the center-surround RF.
Lateral inhibition in the retina occurs as the feedback from the interneurons, horizontal cells and amacrine cells, which receive excitatory inputs from photoreceptors and bipolar cells, respectively, inhibit the excited photoreceptors and bipolar cells and their neighborhood. The typical center-surround RF structure of a ganglion cell is the combination of the excitatory center created by the feedforward cells and inhibitory surround formed by the interneurons. The current study tests the effect of the wide RF component of these interneurons in addition to the classic (narrow) RF (inset).
Fig 3.
Diagrams of Wilson [36] and van Hateren [37,38] model structures and the modifications made to the models in the current study.
Grey box shows the original versions of the models. Modifications made in the current study are shown in green boxes led by the red arrows. See text for details.
Fig 4.
Illustration of the wide RF profiles in the retinal models.
(a) Wilson’s model. The amacrine cell wide RF profile (solid black line) is a modification introduced in this study. The dotted green line shows the horizontal cell filter profile (narrow RF) in the same scale for comparison. (b) The horizontal cell narrow + wide RF in van Hateren’s model (the RF structure is directly taken from [30]). The solid black line illustrates the actual dual-component RF profile and the dotted lines in the plot and the solid lines of the corresponding colors in the inset separately shows the narrow (green) and the wide (red) components for comparison.
Fig 5.
Behavioral and simulation results of Helson’s experiment [11].
(a) Examples of stimuli with narrow bars (left) and wide bars (right). The comparison of stimulus intensity profile with brightness profile schematically illustrates how brightness assimilation and contrast are induced, given narrow bars and wide bars, respectively. (b) The behavioral data on the induction direction and magnitude as a function of grey width for different sizes of bars (replotted from Fig 2 in [11]). A negative / positive induction score indicates assimilation / contrast. (c-h) The retinal model simulation results. (c-e) for Wilson’s model [36] and (f-g) for van Hateren’s model [37,38], with the original algorithms (c, f), modified versions (d, g) and control versions (e, h). All are plotted comparable to (b) except that y-axis is ∆V (mean responses to the grey area among white bars—mean response to the grey area among black bars). Wide RF + parasol spatial processing structures are necessary to predict the bright assimilation-to-contrast induction with increasing bar widths comparable to (b).
Fig 6.
Behavioral and simulation results of Reid and Shapley’s experiment [12].
(a) Examples of stimuli with a narrow ring (left) and a wide ring (right). The comparison of stimulus intensity profile with brightness profile schematically illustrates how a narrower ring induces stronger assimilation compared to a wider ring. (b) The behavioral data on the assimilation strength as a function of difference of the background luminance (labeled ‘induction’ following the terminology in [12]) by different sizes of rings (replotted from Fig 3 in [12]). Larger values indicate stronger assimilation. (c-h) The retinal model simulation results. (c-e) for Wilson’s model and (f-h) for van Hateren’s model, with the original algorithms (c, f), the modified versions (d, g) and the control versions (e, h). All are plotted comparable to (b) except that y-axis is ∆L (the luminance of the disk in the lighter background minus the baseline condition result). The wide RF component is necessary to predict the stronger assimilation magnitude by the narrower rings as in the behavioral data illustrated in (b).
Fig 7.
Simulation results of van Hateren’s model with only the wide RF component of horizontal cells.
(a) The result for Helson’s experiment [11]. (b) The result for Reid and Shapley’s experiment [12]. See and compare with Figs 5G and 6G, respectively.
Fig 8.
Behavioral and simulation results of Rudd and Zemach’s experiment [39].
(a) The behavioral data on the slope of the log match disk luminance vs. log target ring luminance for two subjects, IKZ and JL (see text for details). (b) The simulation results for the same experiment. The green line with ‘+’ markers represent the result of van Hateren’s model and the blue line with ‘o’ markers represent the result of Wilson’s model.