Contributed Talks III: Surface color information gained in searching natural scenes

Area of Science:

• Visual perception
• Ecological psychology
• Computational neuroscience

Background:

• Natural scenes possess complex spatial and spectral characteristics.
• Traditional models of visual search based on object salience may not fully explain gaze behavior in these environments.
• An alternative hypothesis suggests gaze behavior aims to extract information about distinct reflecting elements, characterized by their spectral properties.

Purpose of the Study:

• To investigate whether information gain from eye fixations in natural scenes is related to the spectral diversity of the scene.
• To compare the information obtained from human eye fixations with random sampling.
• To explore the role of reflected color in visual search behavior.

Main Methods:

• Seven observers searched for a neutral target in 20 natural outdoor scenes rendered on a display.
• Eye fixations were recorded, and information was estimated based on reflected spectral radiances and cone photoreceptor excitations.
• Information estimates from fixations were compared to those from randomly sampled points within each scene.

Main Results:

• In 16 out of 20 scenes, eye fixations yielded more information than random sampling.
• The proportion of information gain was largely independent of photoreceptor noise.
• Information gain in each scene positively covaried with color entropy, a measure of spectral diversity.

Conclusions:

• Eye movements in natural scenes may be guided by the principle of maximizing information gain about spectral properties.
• Color entropy, reflecting spectral diversity, is a key factor influencing information gain during visual search.
• Information gain based on spectral characteristics offers a potential foundation for object-oriented explanations of visual search.