Elsevier

Neurocomputing

Volume 69, Issues 10–12, June 2006, Pages 1301-1304
Neurocomputing

A statistical basis for visual field anisotropies

https://doi.org/10.1016/j.neucom.2005.12.096Get rights and content

Abstract

There exist numerous psychophysical paradigms for which performance varies with location of stimulus presentation within the visual field. The following considers potential bases for visual anisotropies, considering the possibility of a statistical basis for such effects in cases where basic sensory asymmetries are present. In particular, the relationship between scene statistics and both upper–lower and lateral visual field asymmetries is considered. Finally, an argument is put forth concerning the apparent radial organization of the visual field, with the suggestion that geometric perspective may give rise to the statistical bias responsible for this effect.

Introduction

It has long been apparent that there exist anisotropies in human visual processing. For example, performance in various psychophysical tasks is much better for grating stimuli oriented horizontally or vertically than for the same stimuli presented at oblique orientations (see [3] for a review). This phenomenon has been termed the oblique effect. Previous efforts have considered a statistical basis for this effect and others, and in the case of the oblique effect, there does exist a bias in image content in favour of vertically and horizontally oriented edges [4]. In this work, we examine a different set of anisotropies, namely, domains for which performance varies as a function of position of stimulus in the visual field. Studies concerning laterality make up the bulk of psychophysical results fitting this category, with performance differences for stimulus presentation in left and right visual field considered. Much of the literature considers the interaction between visual field and spatial frequency of stimuli, with a right visual field advantage for high spatial frequency content and a left visual field advantage for low spatial frequency content. There also exist upper–lower visual field asymmetries that have received relatively less attention in the literature. Articles describing upper–lower visual field asymmetries typically read very similar to a standard visual field laterality study with the exception that the visual world is rotated 90 [1]. It has been suggested that upper–lower visual field asymmetries might arise from the difference in statistics between sky and ground [1]. This claim has yet to be validated through consideration of actual scene statistics. Another interesting anisotropy concerns the so-called radial organization of the visual field. A variety of studies have found that judgments related to line orientation are best for lines oriented towards the centre of the visual field and worst for lines orthogonal to the centre [8], [2]. There does not exist a consensus on the origin of upper–lower and lateral asymmetries, or the so-called radial organization of the visual field. In the sections that follow, each of these effects is considered in the context of local statistics, with the aim of determining whether there might exist a statistical basis for such effects.

Section snippets

A look at the statistics

Explanations for the cause of visual field anisotropies are sparse in the literature, with the majority of work describing what is observed rather than why. The following effort aims at observing the manner in which spatial frequency and orientation statistics vary across the visual field as these are the primary factors considered in most visual field anisotropy studies, and evaluating these observations in the context of existing psychophysical results. In this light, we seek a local

Discussion

We have demonstrated for a variety of visual field anisotropies, that there does appear to be support in the statistics for such effects, especially when such effects apply to very basic paradigms (e.g. detection). Further, we have presented an argument dissociating upper–lower visual field asymmetries from lateral asymmetries, in agreement with more recent psychophysical results. Finally, we have put forth a novel explanation for the apparent radial organization of the visual system,

Neil D.B. Bruce received the degree of Bachelor of Science with a double major in Mathematics and Computer Science from the University of Guelph in 2001. Following this, he attended the University of Waterloo where he received the degree of Master of Applied Science in System Design Engineering. In 2003, he joined the Department of Computer Science and Centre for Vision Research at York University where he is in the process of completing the requirements of the Ph.D. degree. His research

References (9)

There are more references available in the full text version of this article.

Cited by (17)

  • Toward an audiovisual attention model for multimodal video content

    2017, Neurocomputing
    Citation Excerpt :

    Beyond model classes described above, various approaches have been proposed in the literature. One can find models based on the information theory such as Bruce’s approach [44,49,59]. In [46], the authors presented an approach based on the estimation of local saliency.

  • On computational modeling of visual saliency: Examining what's right, and what's left

    2015, Vision Research
    Citation Excerpt :

    There are at least four different sources that are likely to contribute to observed behavior: Anisotropic statistics: Given that the structure of the world (both natural, and manmade), and our typical position within the world (on the ground, with sky or ceiling above, and ground/floor below), there is inherent bias in the statistics of visual input across the visual field (Bruce & Tsotsos, 2006; Nandy & Tjan, 2012). In a more general context, this has implications for the sensitivity to basic features such as edge orientation (van Essen, Newsome, & Maunsell, 1984), or the relative contribution of magnocellular vs. parvocellular pathways (Previc, 1990; McAnany & Levine, 2007) implying both spatial and temporal differences that vary according to position in the visual field.

  • Power spectra of the natural input to the visual system

    2013, Vision Research
    Citation Excerpt :

    The elliptical component tends to be more circularly symmetric at smaller eccentricities, as expected, and gets more elongated with increasing eccentricities. Note that this is an effect of the projective transformation of the visual input and that a related effect is present in perspective distortions in planar projections of visual scenes (Bruce & Tsotsos, 2006). The same procedure was applied to the planar images and the results were similar in terms of orientation and strong activation on the axis, but the values A and α diverge especially in the periphery.

  • Attention selection using global topological properties based on pulse coupled neural network

    2013, Computer Vision and Image Understanding
    Citation Excerpt :

    All participant eye fixations are represented in a binary fixation map. Every image by convolving the binary fixation map with a Gaussian filter [36,41] generates the fixation density map. The fixation density map serves as the ground truth.

  • Orientation anisotropies in macaque visual areas

    2022, Proceedings of the National Academy of Sciences of the United States of America
View all citing articles on Scopus

Neil D.B. Bruce received the degree of Bachelor of Science with a double major in Mathematics and Computer Science from the University of Guelph in 2001. Following this, he attended the University of Waterloo where he received the degree of Master of Applied Science in System Design Engineering. In 2003, he joined the Department of Computer Science and Centre for Vision Research at York University where he is in the process of completing the requirements of the Ph.D. degree. His research interests include computational vision, machine vision, natural image statistics, and machine intelligence.

John K. Tsotsos received an honours undergraduate degree in Engineering Science in 1974 from the University of Toronto and continued at the University of Toronto to complete a Master's degree in 1976 and a Ph.D. in 1980 both in Computer Science. He was a Professor in Computer Science at the University of Toronto from 1980–1999. He is currently at York University where he holds a Tier I Canada Research Chair in Computational Vision, is a Professor in the Department of Computer Science and Engineering and is the Director of York University's Center for Vision Research.

Tsotsos has published many scientific papers, including five conference papers receiving recognition. He has served on numerous conference committees and on the editorial boards of Image and Vision Computing, Computer Vision and Image Understanding, Computational Intelligence and Artificial Intelligence and Medicine.

View full text