Related Experiment Videos
Computational modelling of visual attention
1Hedco Neuroscience Building, University of Southern California, 3641 Watt Way, Los Angeles, California 90089-2520, USA. itti@usc.edu
Nature Reviews. Neuroscience
|March 21, 2001
Summary
Computational models reveal five key trends in visual attention, highlighting context-dependent saliency, saliency maps, and inhibition of return. Attention and eye movements interact, with scene understanding guiding focus.
Area of Science:
- Computational neuroscience
- Cognitive psychology
- Computer vision
Background:
- Focal visual attention is crucial for processing complex scenes.
- Understanding attentional control mechanisms is a key challenge in cognitive science.
Purpose of the Study:
- To outline five emergent trends in computational models of visual attention.
- To provide a framework for understanding the bottom-up control of attentional deployment.
Main Methods:
- Review of recent computational models of focal visual attention.
- Analysis of key trends including saliency, context, inhibition of return, eye movements, and scene understanding.
Main Results:
- Perceptual saliency is context-dependent.
- Saliency maps provide an efficient bottom-up control strategy.
- Inhibition of return is vital for attentional deployment.
- Attention and eye movements exhibit tight interplay.
- Scene understanding constrains attentional selection.
Conclusions:
- These five trends offer a framework for computational and neurobiological insights into visual attention.
- The interplay between image-based features and cognitive factors shapes attentional deployment.