In a series of recent publications, I have proposed a neural model of lightness computation in which large cortical receptive fields spatially integrate the outputs of oriented contrast detectors in V1/V2 (Rudd, 2010, 2013, 2014). The model explains quantitative data on perceptual edge integration in lightness, as well as matching data from the staircase-Gelb and Gilchrist dome paradigms. Here I demonstrate with computer simulations how lightness filling-in results from this same computational model. The filling-in produced by the model differs from that of previous filling-in models in which regions lying between surfaces boundaries are a??colored in.a?? The current model instead envisions filling-in as a a??trans-objecta?? mechanism that supports lightness constancy by establishing a unitary lightness scale that applies to multiple surfaces and objects within the visual scene. I demonstrate how a spreading achromatic color signal that is not stopped by object boundaries can support constancy and at the same time appeara??but only appeara??to be contained by object borders. The effect is achieved through the spatial interaction of spreading of separate lightness and darkness signals that combine like waves to either reinforce or cancel. A spreading darkness signal that encounters a luminance boundary can be partially cancelled by a lightness signal on the other side of the border, thus producing the perceptual impression that the darkness signal was stopped at the border. The model also accounts for Gilchrista??s Area Rule, according to which larger area regions that are not the highest luminance appear lighter than smaller regions having the same luminance. The geometrical patterns of lightness spreading produced by the model depend on the shapes of the model receptive fields. I present simulations produced under different assumptions about these receptive field shapes and explain how they reflect the properties of neural connections within the feedforward ventral pathway of visual cortex.
展开▼
