Highly informative natural scene regions increase microsaccade production during visual scanning

Area of Science:

• Vision science
• Computational neuroscience
• Human-computer interaction

Background:

• Classical image statistics (e.g., contrast, entropy) are traditionally linked to ocular fixation targeting.
• These statistics may not fully capture task-relevant informativeness or the role of microsaccades in information acquisition.
• Previous measures of informativeness lacked objectivity and consistency.

Purpose of the Study:

• To develop an objective measure of image region informativeness.
• To investigate the relationship between image informativeness, classical image statistics, and microsaccade production.
• To understand the role of microsaccades in visual information acquisition.

Main Methods:

• Developed an objective informativeness measure based on human fixation consistency.
• Analyzed microsaccade rates in image regions with varying informativeness and classical statistics.
• Controlled for fixation duration to isolate the effect of microsaccades.

Main Results:

• More microsaccades were generated in highly informative image regions compared to less informative ones.
• Highly informative regions showed lower redundancy in classical image statistics.
• Increased microsaccade production was independent of fixation duration.

Conclusions:

• Microsaccades are actively employed by the visual system to enhance information acquisition from salient image regions.
• The developed objective informativeness measure provides a more consistent and task-relevant approach.
• Microsaccade production is linked to information-seeking behavior beyond simple fixation guidance.