Eye movement statistics in humans are consistent with an optimal search strategy
Jiri Najemnik1, Wilson S Geisler
1Center for Perceptual Systems and Department of Psychology, University of Texas at Austin, TX 78712-0187, USA. najemnik@mail.cps.utexas.edu
Journal of Vision
|May 20, 2008
Summary
Human visual search strategies were compared. Both humans and ideal searchers focused on specific areas, unlike MAP searchers, suggesting information gain drives sophisticated visual search mechanisms.
Area of Science:
- Cognitive Psychology
- Computational Neuroscience
- Vision Science
Background:
- Visual search models often assume feature matching guides eye movements.
- Two key strategies are maximum a posteriori (MAP) search and ideal search, which maximizes information gain.
Purpose of the Study:
- To compare human eye movements with MAP and ideal search strategies.
- To investigate how humans conduct visual search in naturalistic scenes.
Main Methods:
- Comparing human eye movements to computational models (MAP and ideal search).
- Utilizing tasks with known targets in random backgrounds mimicking natural scenes.
- Analyzing fixation patterns in a circular search area.
Main Results:
- Human and ideal searchers showed similar fixation patterns, concentrating around a central "donut" region with high density at the top and bottom.
- MAP searchers exhibited a more uniform fixation distribution, with low density at the top and bottom.
- This contrasts with the uniform distribution predicted by MAP search.
Conclusions:
- Human visual search is not solely based on feature matching (MAP).
- The findings support a sophisticated search mechanism that prioritizes maximizing information gain across fixations.
- Eye movement patterns suggest a strategy that optimizes information acquisition over simple feature correlation.
