Week 1: Cog Forum - (3:30 PM - 5:00 PM) 10/07/05
Title: Perception of Visual Structure: A Boolean Map Theory
Speaker: Hal Pashler
From: UCSD
Abstract: Studies of the perception of multidimensional displays have focused almost entirely on search tasks. However, observers often view scenes without trying to find any particular target, and when doing so, they frequently apprehend structures and patterns, not just the presence of "target" elements. Exactly what kinds of structures and patterns can people detect without scrutiny, and what role does attention play? We propose and discuss what we call the boolean mapping theory, according to which perception of structure in multidimensional displays is constrained to occur through one or more iterations of a two-stage sequence: (1) voluntary dichotomization of the display by attending to a subregion in a single featural dimension, and (2) the spatial analysis of the "positive" regions of the display. This hypothesis is related to feature integration theory, but differs in some important ways. Some tests of the theory will be presented, involving perception of figure-ground segregation, color symmetry, and various forms of clustering and higher order structures. The practical implications for visual interpretation of multidimensional data (e.g., in visual data mining and medical imaging) will also be touched upon. (Note: This work was done in collaboration with Liqiang Huang).
-----
-----
A question from Prof. Robert A. Bjork,
" Then, what is map? "
Title: Perception of Visual Structure: A Boolean Map Theory
Speaker: Hal Pashler
From: UCSD
Abstract: Studies of the perception of multidimensional displays have focused almost entirely on search tasks. However, observers often view scenes without trying to find any particular target, and when doing so, they frequently apprehend structures and patterns, not just the presence of "target" elements. Exactly what kinds of structures and patterns can people detect without scrutiny, and what role does attention play? We propose and discuss what we call the boolean mapping theory, according to which perception of structure in multidimensional displays is constrained to occur through one or more iterations of a two-stage sequence: (1) voluntary dichotomization of the display by attending to a subregion in a single featural dimension, and (2) the spatial analysis of the "positive" regions of the display. This hypothesis is related to feature integration theory, but differs in some important ways. Some tests of the theory will be presented, involving perception of figure-ground segregation, color symmetry, and various forms of clustering and higher order structures. The practical implications for visual interpretation of multidimensional data (e.g., in visual data mining and medical imaging) will also be touched upon. (Note: This work was done in collaboration with Liqiang Huang).
-----
-----
A question from Prof. Robert A. Bjork,
" Then, what is map? "
アクセス
トータル
ランキング
