Center-surround patterns emerge as optimal predictors for human saccade targets

Area of Science:

• Neuroscience
• Computational Vision
• Visual Perception

Background:

• The human visual system is foveated, leading to rapid declines in resolution and acuity outside the central visual field.
• Despite foveation, significant portions of visual scenes are processed rapidly via saccadic eye movements, indicating efficient target selection strategies.
• Local image structure at saccade targets is known to influence selection, but the specific relevant features remain debated.

Purpose of the Study:

• To identify the most relevant visual features for predicting saccade targets.
• To develop and test a computational model for saccade target selection based on local image structure.
• To compare the predictive power of a novel model with existing complex models.

Main Methods:

• Development of a computational model based on center-surround patterns to predict saccade targets.
• Evaluation of the model's predictive accuracy using image structure data.
• Comparison of the model's performance against previously proposed, more complex models.

Main Results:

• Center-surround patterns were identified as the optimal solution for predicting saccade targets from local image structure.
• A simple, one-layer feed-forward network model based on these patterns demonstrated high predictive accuracy.
• The model's performance was comparable to more complex models involving multi-scale processing and multiple feature channels.

Conclusions:

• Simple center-surround patterns are highly effective in predicting saccade targets.
• The findings suggest that bottom-up visual saliency may be computed sub-cortically, potentially in structures like the superior colliculus, rather than primarily in the cortex.
• This challenges previous assumptions about the complexity of visual attention mechanisms.