Project

Back to overview

What information defines an object before it is selected and recognized?

English title What information defines an object before it is selected and recognized?
Applicant Krummenacher Joseph
Number 130252
Funding scheme Project funding (Div. I-III)
Research institution Département de Psychologie Université de Fribourg
Institution of higher education University of Fribourg - FR
Main discipline Psychology
Start/End 01.04.2010 - 31.03.2012
Approved amount 146'963.00
Show all

Keywords (10)

perception; selection; visual representation; proto-object; feature-based attention; saliency-based attention; visual perception; visual features; feature-based selection; topological characteristics

Lay Summary (English)

Lead
Lay summary
It was shown that visual attention can be directed to objects. As an example, Duncan (1984) showed that participants were less accurate in reporting two features of different objects compared with two features of a single object. Findings such as Duncan's imply a coarse object representation that is generated prior to the deployment of selective attention. Early object representations are referred to as "proto-objects". The notion of proto-objects is accepted, but the exact nature of the representation and the mechanisms underlying the initial segmentation of visual input into proto-objects are not understood at present. Classical theories of perception assume that proto-objects are assembled from feature representations (color, orientation, size). Models include Treisman et al.'s (1980) Feature Integration Theory and Wolfe's (1994) Guided Search model. Feature-based accounts are supported by behavioral, psycho- and neuro-physiological evidence.The Gestalt psychologists demonstrated that human perception is sensitive to characteristics of the configuration of objects rather than to mere features. The visual system seems to segment scenes into grouped regions, proto-objects, before initiating search and recognition processes for object identification. However, Gestalt principles (e.g., proximity, similarity, closure) are descriptive rather than explanatory and post-hoc rather than predictive and open to scientific examination. In an attempt to avoid the problems of Gestalt accounts, Chen (e.g., 1982) using psychophysical methods, has emphasized the importance of topological properties for the initial visual perceptual organization. The term topology is used to describe relationships between object properties as connectivity, existence of holes or inside/outside relationships. Chen (2005) proposed that the generation of proto-objects is based on topological characteristics. The topological account constitutes a challenge to key concepts of the feature-based accounts; in particular it reverses the sequence of events of the component modules of visual information processing.The main objective of the present research project is to examine information that underlies structuring of visual scenes. Four experimental series using psychophysical methods were designed to differentiate between the two theoretical positions and to contribute to the understanding of the mechanisms underlying visual selection and the generation of an initial representation of a visual scene.
Direct link to Lay Summary Last update: 21.02.2013

Responsible applicant and co-applicants

Employees

Associated projects

Number Title Start Funding scheme
110543 Selective visual processing: An integrative neurocognitive account based on psychological research methods 01.09.2006 SNSF Professorships

Abstract

Visual attention can be directed to objects (Goldsmith, 1998). This implies that some sort of coarse object representation is generated from the visual input prior to the deployment of attention and object recognition. We will call these representations “proto-objects” (Rensink, 2000; Wolfe & Bennett, 1997). However, the exact nature and the mechanisms underlying this initial segmentation of visual input into proto-objects or proto-representations are not understood at present. The main aim of the present research project is to examine the type of information that underlies structuring. Chen (1982, 1990, 2005a; Zhou, Chen, & Zhang, 1992) has emphasized the importance of topological properties (such as the connectivity and number of holes) for processes of initial visual perceptual organization. Chen argues that perception proceeds in a “holistic”, global-to-local manner (see also Hochstein & Ahissar, 2002); stated differently, the perception of a face with two eyes and a mouth occurs prior to the perception of local properties such as the shape, size and orientation of the face. This approach is in strong contrast to the local-to-global approach that has been dominant in theoretical approaches to visual perception and information selection and assumes that the generation of a coherent object representation requires the integration of a limited set of features such as color, orientation, or size (Treisman and Gelade, 1980; Wolfe, 1994, 2005). The methodological approach by Chen (2005a) proposes that the generation of proto-objects is based on topological characteristics of visual information. This assumption challenges key concepts of feature-based accounts. In the present research project predictions of the two theoretical positions regarding early mechanisms of visual selection will be examined; in particular the project is aimed at investigating whether (global) topological features or local features assume the central role (a) in the initial segmentation of a visual scene (image) into proto-objects (research question 1), and (b) in the subsequent guidance of selective attention to those proto-objects (research question 2). Further, the questions whether dynamic proto-objects must keep their topological structure (research question 3) in order to be traceable; and the role of topology for the perception of wholes and parts will be investigated (research question 4).These research questions will be tested in using the “visual search” as well as the “multiple object tracking” paradigms - both constitute more realistic paradigms than the two-object matching task used by Chen and colleagues. It is expected that the results of the four planned experimental series will be instrumental in differentiating between the two theoretical positions.Overall, the results are expected to contribute significantly to the detailed understanding of the mechanisms underlying visual selection and the generation of an initial structure of a visual scene.
-