Abstract
Analyzing the time course of eye movements during free exploration of real-world scenes often reveal an increase in fixation durations together with a decrease in saccade amplitudes which has been explained as a shift from ambient (global) to focal (local) processing. The ambient mode refers to bottom-up processing, serving global orientation controlled by the saliency of the stimulus. In contrast, the focal mode is rather related to top-down processing and associated with the identification of object features. The ambient-to-focal strategy seems to be evoked by the onset of images but a systematic investigation for dynamic stimuli is still missing. If this pattern represents an omnipresent viewing characteristic for new visual information, it should also be evoked by the onset of dynamic scenes. Furthermore, an interruption of the dynamic information flow, caused by a scene change, might also lead to a restart of the ambient-to-focal strategy. To address this question, we recorded videos where the viewer walked through indoor scenarios when no other persons were present. From the recorded material we generated 3 different video clips, each with an approximate length of 10 minutes, containing several scene changes. Scene changes were defined as either abrupt scene cuts (e.g. cut from room to staircase) or pan shots (horizontal camera scans providing a new perspective). We analyzed fixation durations and saccade amplitudes focusing on the time course of visual exploration following the video onset, differences before and after scene changes, and spatio-temporal characteristics of the viewing behavior in relation to the stimulus saliency and the scanning similarity between subjects. Initial results indicate the ambient-to-focal strategy also for the exploration of dynamic stimuli. These findings further support the ambient/focal distinction, which can help to decode complex viewing pattern into distinct processing levels that are related to the interplay of bottom-up and top-down mechanisms.
Meeting abstract presented at VSS 2014