Abstract
The visual system is constrained by limitations that challenge the efficient and unambiguous encoding of commonly encountered signals. To understand how human vision overcomes these hurdles, we must consider its sampling method (top-down and bottom-up guided saccades and fixations) as well as the statistics of the visual environment. Previous studies have shown that image statistics can influence fixation locations, but their effects on the production of microsaccades during fixation are unknown. Indeed, no research has shown a role of microsaccades in information acquisition during visual scanning. Microsaccade production and information acquisition could be linked in a variety of ways: 1) Microsaccades may sample image regions where information content is high. If so, more informative regions should trigger higher microsaccade rates than less informative regions; 2) Microsaccades may instead extract information from image regions where information content is low. If so, less informative regions should trigger higher microsaccade rates; 3) Microsaccades may not be related to information acquisition, in which case microsaccade rates should be equivalent in more and less informative regions. To determine which possibility is correct, we must first specify more versus less informative image regions. We defined more versus less informative regions in terms of fixation consistency across observers, so as to account for both bottom-up and top-down influences on image exploration. We then compared the characteristics of fixations in more versus less informative regions. We also analyzed the classical statistics of image patches around fixations as a function of informativeness and microsaccade production. Viewers generated more microsaccades and fixated for longer periods on more informative image regions. Such regions were less redundant in terms of their contrast, entropy, and correlation. Further, microsaccade production was not fully explained by fixation duration, suggesting that the visual system specifically employs microsaccades to heighten information acquisition from highly informative image regions.
Meeting abstract presented at VSS 2014