August 2014
Volume 14, Issue 10
Vision Sciences Society Annual Meeting Abstract  |   August 2014
Rapidly estimating numerosity independent of size-related distance or occlusion
Author Affiliations
  • Guillaume Riesen
    Computational Vision Laboratory, Northeastern University
  • Harald Ruda
    Computational Vision Laboratory, Northeastern University
  • Ennio Mingolla
    Computational Vision Laboratory, Northeastern University
Journal of Vision August 2014, Vol.14, 879. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Guillaume Riesen, Harald Ruda, Ennio Mingolla; Rapidly estimating numerosity independent of size-related distance or occlusion. Journal of Vision 2014;14(10):879.

      Download citation file:

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

  • Supplements

Finding objects in natural environments presents a challenge for even simple objects with known properties. Because of occlusion and distance, the number and extent (visual angle) of visible surfaces do not directly convey the number of objects seen. Are those three reddish surfaces one nearby partially occluded apple or three distant apples? Resolving such ambiguities seems effortless for human observers. To benchmark human performance, we asked subjects to indicate which of two briefly pictured trees had more apples. Trees were generated in a virtual environment, and pairs were constructed with variations in occlusion, distance, or number of visible apples. Preliminary results show good performance (>95% correct) on the task, despite some less fruit-laden trees having double the visible apple area of their partners. Subjects did not simply respond to the extent of visible apple surfaces; their judgments were robust to variations in occlusion or visual angle. We believe this is the result of a multi-stage process similar to Xu and Chun's (TiCS, 2009) "individuation and identification" stages, where proto-objects with low-level features are first spatially localized then further refined into identifiable objects. We created a simple algorithm to investigate this hypothesis. First, it segments pixels with "tree-like" color (RGB values) by blurring and thresholding. It then tags apple-colored pixels within these tree-like regions. These pixels are convolved with a circular template whose radius depends on the area of the tree-like region. Local maximums give locations of apples that might explain the observed pixels, given the estimated tree size. This system can distinguish between a cluster of patches representing a single heavily-occluded apple nearby, versus multiple apples at a distance. It performs comparably to humans on our task, suggesting that subjects group proto-objects derived from visible patches into individual apples as a basis for numerosity judgments.

Meeting abstract presented at VSS 2014


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.