December 2022
Volume 22, Issue 14
Open Access
Vision Sciences Society Annual Meeting Abstract  |   December 2022
Locally available achromatic cues can reliably distinguish occlusion boundaries from cast shadows
Author Affiliations & Notes
  • Christopher DiMattina
    Computational Perception Laboratory
    Department of Psychology, Florida Gulf Coast University
    FGCU Computational Facility
  • Lauren Anderson
    Computational Perception Laboratory
    Department of Psychology, Florida Gulf Coast University
  • Josiah Burnham
    Computational Perception Laboratory
    Department of Software Engineering, Florida Gulf Coast University
  • Michelle DeAngelis
    Computational Perception Laboratory
    Department of Psychology, Florida Gulf Coast University
  • Betul Guner
    Computational Perception Laboratory
    Department of Psychology, Florida Gulf Coast University
  • Footnotes
    Acknowledgements  Funded by NIH-R15-EY032732-01 to C.D.
Journal of Vision December 2022, Vol.22, 3039. doi:https://doi.org/10.1167/jov.22.14.3039
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Christopher DiMattina, Lauren Anderson, Josiah Burnham, Michelle DeAngelis, Betul Guner; Locally available achromatic cues can reliably distinguish occlusion boundaries from cast shadows. Journal of Vision 2022;22(14):3039. https://doi.org/10.1167/jov.22.14.3039.

      Download citation file:


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

      ×
  • Supplements
Abstract

Luminance edges in natural scenes can arise from physical causes which give rise to different scene interpretations: Surface boundaries, changes in reflectance, and cast shadows all give rise to luminance edges. In order to better understand the locally available achromatic cues which distinguish cast shadows from occlusion boundaries, we developed a large database of hand-labeled shadow contours using a set of N = 47 images taken from the McGill Calibrated Color Image Database. We compared the statistical properties of these shadows to statistics measured from two different sets of occlusion contours also derived from the McGill Database. We found that RMS and Michelson contrasts measured from occlusions spanned a wider range than these same quantities measured from shadows. Occlusions accounted for both the highest and lowest observed contrasts, extending earlier observations of higher, but not lower, contrasts for occlusions. We suggest that our observation of many low-contrast occlusions is due to our large corpus of hand-labeled occlusion images which transcend the limitations of datasets used previously. In addition to measuring contrast statistics, we also demonstrated that shadows and occlusions can be distinguished by differences in spatial frequency content, with occlusions having a larger proportion of their energy in the high spatial frequency range than shadows. Finally, we evaluated the ability of a multiscale bank of oriented Gabor filters resembling V1 simple cells to classify these two categories of edges. We implemented binomial logistic regression on the outputs of a multiscale Gabor filter bank, and found this simple model could reach nearly 80% accuracy on the task of discriminating these categories at a resolution of 40x40 pixels. Taken as a whole, our results show us that an essential cue for global scene interpretation like the cause of a luminance edge can be partially disambiguated using locally available information.

×
×

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.

×