Scene Content Selected by Active Vision
The primate visual system actively selects visual information
from the environment for detailed processing through mechanisms of
visual attention and saccadic eye movements. This study examines the
statistical properties of the scene content selected by active vision.
Eye movements were recorded while participants free-viewed digitized
images of natural and artificial scenes. Fixation locations were
determined for each image and image patches were extracted around the
observed fixation locations. Measures of local contrast, local spatial
correlation and spatial frequency content were calculated on the
extracted image patches. Replicating previous results, local contrast
was found to be greater at the points of fixation when compared to
either the contrast for image patches extracted at random locations or
at the observed fixation locations using an image-shuffled
database. Contrary to some results and in agreement with other
results in the literature, a significant decorrelation of image intensity
is observed between the locations of fixation and other neighboring
locations. A discussion and analysis of methodological techniques is
given that provides an explanation for the discrepancy in results. The
results of our analyses indicate that both the local contrast
and correlation at the points of fixation are a function of image type
and, furthermore, that the magnitude of these effects depend on the
levels of contrast and correlation present overall in the
images. Finally, the largest effect sizes in local contrast and
correlation are found at distances of approximately 1 degree of visual
angle, which agrees well with measures of optimal spatial scale
selectivity in the visual periphery where visual information for
potential saccade targets is processed.
The following paper describes this research:
Parkhurst, D. J., and Niebur, E. (2003). Scene content selected by active vision. Spatial
Vision, 6(2), 125-154.
Stimulus-Driven Visual Selective Attention
We are interested in how attention is allocated when people view complex
natural scenes. We used a biologically-motivated computational model
of stimulus-driven visual selective attention to measure the degree
to which attentional allocation in such conditions is dependent on
the stimulus properties. We recorded eye movements of human participants
freeviewing images of natural scenes as a measure of attentional
allocation. A good correspondance between stimlus salience and the
observed eye movements was observed. The following paper describes
the results of this research in detail:
Parkhurst, Law, and Niebur (2002). Modeling the role of salience in the
allocation of overt visual attention. Vision Research, 42(1), 107-123.
People involved:
Derrick Parkhurst
Klinton Law
Variable Resolution Displays
Gaze-contingent variable resolution displays present a high level of
detail at the viewer's point of gaze and lower resolution in the
periphery. Variable resolution techniques take advantage of the well
known fact that the human visual system sensitivity is greatest at the
point of gaze and rapidly falls off in the periphery. By rendering
detail only where it can be processed effectively, computational
resources, that are typically wasted displaying "unseen" detail in the
periphery, are saved and can be used for other purposes.
Applications of variable resolution displays range from virtual reality
to internet image transmission. For example, in virtual reality, many
resources are required to render a detailed virtual scene. Using a
variable resolution display allows otherwise wasted
computational resources to be shifted to increase frame update rates.
In internet image transmission applications, communication bandwidths
are often limited and varialbe resolution displays can be used to save
on communication resources by transmitting only the most important
information (that at the point of gaze, or that which the user
specifies).
We've investigated the behavioral consequences of using a
gaze-contingent variable resolution display with a visual
search paradigm. Although there can be dramatic behavioral consequences
(e.g. altered search performance, abnormal eye movements, etc.) under
the appropriate conditions, these effects can be minimized or
eliminated. Our results are described in the following paper:
Parkhurst, D., Culurciello, E., & Niebur, E. (2000). Evaluating
variable resolution diplays with visual search: Task performance and
eye movements. Proceedings of the ACM Eye Tracking Reseach and
Applications Symposium, 1, 105-109.
We've also conducted a review of the behavioral literature, evaluated practical
constraints and conducted a theoretical analysis of the potential computational
savings that variable resolutions displays can achieve. We've integrated these
issues into a recent review paper:
Parkhurst, D. and Niebur, E. (2002). Variable-Resolution Displays: A theoretical,
practical and behavioral evaluation.Human Factors, 44(4), 611--29.
People involved in this project:
Derrick Parkhurst
Eugenio Culurciello
Ernst Niebur
Gaze-Contingent LOD Rendering in Virtual Reality
Velocity-based Level of Detail Rendering
We are currently investigating the behavioral consequences of using
velocity-based variable resolution display techniques in virtual
reality. We've adapted the Unreal
rendering engine to display meshes at various levels of detail, depending
on the velocity of the objects across the visual field. Given that visual
sensitivity to the details of moving objects is significantly reduced,
computation resources can be shifted to increase frame rates
by rendering moving objects in less detail. Many thanks go out to the
generous support of those
at Epic Games for making this research possible.
A recent unpublished lab technical report describes our efforts in this
project:
Parkhurst, D., and Niebur, E. (2001).
Evaluating velocity-based level of detail rendering of virtual environments
using visual search.
In Lab Technical Report 2001-01, pgs. 1--6. [PDF]
Gaze-Contingent Level of Detail Rendering
We are currently investigating the behavioral consequences of using
gaze-contingent variable resolution display techniques in virtual
reality. We've adapted the Unreal
rendering engine to display meshes at various levels of detail, depending
on the user's point of gaze. Significant savings are seen in the way of
increased frame rates using this technique because "unseen" detail in the
visual periphery is not rendered. Again, many thanks go out to the
generous support of those at Epic
Games for making this research possible.
A recent unpublished lab technical report describes our efforts in this
project:
Parkhurst, D., Law, I., and Niebur, E. (2001).
Evaluating gaze-contingent level of detail rendering of virtual environments
using visual search.
In Lab Technical Report 2001-02, pgs. 1--6. [PDF]
People involved:
Derrick Parkhurst
Irwin Law
Ernst Niebur
Go Back To Computational Neuroscience Lab