How do image complexity, task demands and looking biases influence human gaze behavior?

In this paper we propose an information-theoretic approach to understand eye-movement patterns, in relation to the task performed and image complexity. We commence with the analysis of the distributions and amplitudes of eye-movement saccades, performed across two different image-viewing tasks: free viewing and visual search. Our working hypothesis is that the complexity of image information and task demands should interact. This should be reflected in the Markovian pattern of short and long saccades. We compute high-order Markovian models of performing a large saccade after many short ones and also propose a novel method for quantifying image complexity. The analysis of the interaction between high-order Markovianity, task and image complexity supports our hypothesis.

► Image statistics and task demands have a joint influence on unrestrained gaze behavior.
► Novel information-theoretical algorithm for image complexity.
► Image complexity metric discriminates between task-dependent gaze behavioral patterns.
► High-order Markovianity in saccadic behavior also depends on image complexity.