September 2015
Volume 15, Issue 12
Vision Sciences Society Annual Meeting Abstract  |   September 2015
Effects of Image Size on Clutter Perception: More Evidence for Proto-Object Segmentation
Author Affiliations
  • Chen-Ping Yu
    Department of Computer Science, Stony Brook University
  • Gregory Zelinsky
    Department of Computer Science, Stony Brook University Department of Psychology, Stony Brook University
Journal of Vision September 2015, Vol.15, 120. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Chen-Ping Yu, Gregory Zelinsky; Effects of Image Size on Clutter Perception: More Evidence for Proto-Object Segmentation. Journal of Vision 2015;15(12):120.

      Download citation file:

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

  • Supplements

Outside the context of search, little is known about visual clutter perception. This study asked how clutter perception is affected by retinal image size. Following Yu, Samaras, and Zelinsky (2014, JoV), Experiment 1 had participants (n=20) rank order 90 images of random-category realistic scenes from least-to-most cluttered. Scenes were smaller (6.75°×5°) versions of the larger (27°×20°) images from the earlier study. Comparing the median clutter ranking (aggregated over participants) to the median ranking reported for the larger scenes yielded a near perfect correlation (Spearman's ρ = .953, p < .001), suggesting that relative clutter perception is invariant to image size. Experiment 2 had new participants (n=16) again rank order the same 90 images, now evenly divided between small and large (counterbalanced over participants). After obtaining difference scores (small image ranking minus large image ranking) for all 90 scenes we found that smaller images tended to be ranked as more cluttered than larger images (Wilcoxon signed-rank test, p < 0.001), suggesting that absolute clutter perception increases with decreasing image size. Clutter was estimated for each of the 90 small and large scenes using the proto-object model of clutter perception. This model segments an image into superpixels, then merges neighboring superpixels that share a common color cluster to form proto-objects. Clutter estimates were obtained by counting the proto-objects in each scene, and these were rank ordered from least-to-most cluttered. Correlating this ranking with the behavioral ranking from Experiment 1 yielded a Spearman's ρ =.852 (p < .001). We also compared the number of proto-objects in small and large scenes using the method from Experiment 2 and found that smaller scenes were estimated to be more cluttered than their larger counterparts (p < 0.001). We conclude that clutter perception is mediated by proto-objects, a mid-level of visual representation between features and objects.

Meeting abstract presented at VSS 2015


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.