Figure—Ground Segregation in a Recurrent Network Architecture

Abstract

Figure-ground segregation in a recurrent network architectureLamme, V.A.F.; Roelfsema, P.R.; Spekreijse, H.; Bosch, H. Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: http://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. Abstract& Here we propose a model of how the visual brain segregates textured scenes into figures and background. During texture segregation, locations where the properties of texture elements change abruptly are assigned to boundaries, whereas image regions that are relatively homogeneous are grouped together. Boundary detection and grouping of image regions require different connection schemes, which are accommodated in a single network architecture by implementing them in different layers. As a result, all units carry signals related to boundary detection as well as grouping of image regions, in accordance with cortical physiology. Boundaries yield an early enhancement of network responses, but at a later point, an entire figural region is grouped together, because units that respond to it are labeled with enhanced activity. The model predicts which image regions are preferentially perceived as figure or as background and reproduces the spatio-temporal profile of neuronal activity in the visual cortex during texture segregation in intact animals, as well as in animals with cortical lesions. &

Neural evidence for competition-mediated suppression in the perception of a single object

“…In these models, units on the winning side (the object or "figure" side) are ultimately enhanced relative to units on the losing side (the groundside). And indeed, neural evidence shows that responses to figures are enhanced relative to responses to grounds (Roelfsema et al, 2002).…”

“…4 object (or figure); the other side is perceived as a locally shapeless ground continuing behind the object. Computational theories of figure-ground perception implement suppressive competition between low-level edge-units, feature-units and/or image-based properties such as convexity, symmetry, and small area detected on opposite sides of a shared border (e.g., Craft, Schutze, Niebur, & von der Heydt, 2007;Grossberg, 1994;Kienker, Sejnowski, & Hinton, 1986;Kogo, Strecha, Van Gool, & Wagemans, 2010;Roelfsema, Lamme, Spekreijse, & Bosch, 2002;Sejnowski & Hinton, 1987;Vecera & O'Reilly, 1998). In these models, units on the winning side (the object or "figure" side) are ultimately enhanced relative to units on the losing side (the groundside).…”

Act Quickly, Decide Later: Long-latency Visual Processing Underlies Perceptual Decisions but Not Reflexive Behavior

“…Visual processing of texture checkerboards requires texture segregation, which is a two-stage process: First, borders of figures are detected, and subsequently, the figures are filled in. Border detection occurs around 80-90 msec, and is likely to be the result of lateral inhibition within cortical areas, whereas figure filling-in can take up to 200 msec and depends on re-entrant processing (Scholte, Jolij, Fahrenfort, & Lamme, 2008;Jehee, Roelfsema, Deco, Murre, & Lamme, 2007;Heinen et al, 2005;Roelfsema, Lamme, Spekreijse, & Bosch, 2002;Caputo & Casco, 1999;Lamme, 1995). This latter stage has been linked to perceptual awareness of texture stimuli, whereas the former stage could be sufficient in order to detect presence of a texture stimulus Heinen et al, 2005;Lamme, 1995Lamme, , 2003Supèr, Lamme, & Spekreijse, 2001).…”

Occipital network for figure/ground organization

“…What neural architecture allows local and global information to be combined rapidly, to provide for Wgure/ground categorization? In terms of neural architecture, does the mechanism employ lateral interactions (e.g., Gerrits and Vendrik 1970;Grossberg and Mingolla 1985;Grossberg 1994;Baek and Sajda 2005;Pao et al 1999;Zhaoping 2005), top-down feedback (e.g., Lamme and Roelfsema 2000;Lamme et al 2002;Roelfsema et al 2002;Craft et al 2007) or more sophisticated combinations? In terms of its coding mechanism, do the Wgure/ground conWgurations use border-based coding (e.g., von der Heydt et al 2003Heydt et al , 2005Craft et al 2007) or region-based coding (e.g., Lamme 1995;Zipser et al 1996;Lamme et al 1998;Super et al 2001;Roelfsema et al 2002)?…”