Geometric-Optical Illusions from Early Visual Cortex: Computational Principles and Top-Down Feedback

• Applicant: Junhao Liang, Ph.D.
• Subject Area: Biological Psychology and Cognitive Neuroscience ; Human Cognitive and Systems Neuroscience
• Term: from 2021 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 465358224

• Final Report Year: 2023

Final Report Abstract

The project planned to use mathematical model to study the neural dynamics of early visual cortex and use human visual psychophysics experiments to test the perceptual difference between central versus peripheral vision on geometric-optical illusions, aiming to probe their underlying neural mechanism. First, we built an image-computable model of primary visual cortex (V1). Capturing the multiscale orientation encoding properties of V1, the model processes visual inputs from a multiscale steerable pyramid decomposition from a 2D image. These visual inputs are modulated by neural lateral interactions in the model, capturing the essential neurophysiological structure of the visual cortex. The modulated outputs are used to reconstruct a new image. Thus, the model mimics the pre-attentive modulation by V1 mechanism. Second, we conducted a human visual psychophysics experiment with eye-tracking on a visual search task inspired by a flip tilt illusion featuring the difference between central and peripheral vision. Our finding suggested the role of V1 in computing the bottom-up saliency of visual inputs and suggesting a central-peripheral dichotomy in human vision that central and peripheral vision are specialized for visual decoding and visual selection, respectively. These results from the project provide novel insights into the mechanism of V1 on visual information processing and inspire further studies in spatial vision.

Publications

• Further evidence for V1 Saliency Hypothesis from an anomalous visual search behavior: a replication study and its extension. Bernstein Conference 2021.
  Maher, S.; Riedel, C.; Karlej, J.; Liang, J. & Zhaoping, L.
  
• Criticality enhances the multilevel reliability of stimulus responses in cortical neural networks. PLOS Computational Biology, 18(1), e1009848.
  Liang, Junhao & Zhou, Changsong
  
• Eye tracking evidence for V1 Saliency Hypothesis from an anomalous visual search behavior. Annual Meeting of the Vision Sciences Society (V-VSS 2022), St. Pete Beach, FL, USA.
  Liang, J.; Maher, S. & Zhaoping, L.
  
• Perceptually non-distinctive uniqueness in eye of origin of visual inputs boosts saliency by unique color and/or orientation: implications for mechanisms in the primary visual cortex. European Conference on Visual Perception (ECVP 2022), Nijmegen, the Netherlands.
  Zou, J.; Liang, J. & Zhaoping, L.
  
• Saliency by barely visible uniqueness in eye of origin of visual inputs: its assessment by the V1 Saliency Hypothesis and its interaction with saliency by color and/or orientation. Bernstein Conference 2022., Berlin, Germany.
  Zou, J.; Liang, J. & Zhaoping, L.
  
• Eye movement evidence for the V1 Saliency Hypothesis and the Central-peripheral Dichotomy theory in an anomalous visual search task. Vision Research, 212, 108308.
  Liang, Junhao; Maher, Severin & Zhaoping, Li