June 2004
Volume 4, Issue 8
Free
Vision Sciences Society Annual Meeting Abstract  |   August 2004
Bottom-up and top-down influences on attentional allocation in natural dynamic scenes
Author Affiliations
  • Ran Carmi
    Neuroscience Graduate Program, University of Southern California, USA
  • Laurent Itti
    Departments of Computer Science, Psychology and Neuroscience Graduate Program, University of Southern California, USA
Journal of Vision August 2004, Vol.4, 864. doi:10.1167/4.8.864
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Ran Carmi, Laurent Itti; Bottom-up and top-down influences on attentional allocation in natural dynamic scenes. Journal of Vision 2004;4(8):864. doi: 10.1167/4.8.864.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Motivation: Attentional allocation is often studied by isolating a small subset of bottom-up or top-down influences in highly-controlled environments. Other studies provide phenomenological descriptions of overt attention in highly-complex environments (e.g., Land & McLeod 2000, Nat. Neuro.). The present study attempts to find the middle ground by investigating the temporal interplay between bottom-up and top-down influences on attentional allocation in dynamic natural scenes. Methods: A set of heterogeneous video clips was cut into clippets (M=2s), which were scrambled and reassembled into MTV-style clips. Subjects were instructed to “follow the main actors and actions”. Eye positions were recorded using an eye-tracker, yielding a total of 1.35 million samples, which were segmented into saccades, blinks, and fixation/smooth pursuit periods. Saccade target selection was compared to predictions made by a computational model of saliency-based attention capture (Itti & Koch 2000, Vis. Res.). Results were compared to those of a twin experiment that employed the same methodology, but used the unscrambled clips as stimuli. Results: The ratio average/max saliency calculated by the model at human saccade targets was 0.47 compared with 0.25 at random saccade targets. The null hypothesis that human and random samples were drawn from the same distribution was rejected by sign and wilcoxon tests (p<10E-10 for both). Identical results were obtained for both the scrambled and unscrambled stimuli. Conclusions: Eliminating visual context effects beyond the first ∼2s by temporally scrambling dynamic natural scenes did not increase the relative weight of bottom-up influences on attentional allocation. A possible interpretation of our results is that the relative weight of top-down influences on attentional allocation in natural viewing conditions does not change with presentation time (at least beyond the first ∼2s).

Carmi, R., Itti, L.(2004). Bottom-up and top-down influences on attentional allocation in natural dynamic scenes [Abstract]. Journal of Vision, 4( 8): 864, 864a, http://journalofvision.org/4/8/864/, doi:10.1167/4.8.864. [CrossRef]
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×