Misperception of rigidity from actively generated optic flow.

Area of Science:

• Visual perception
• Computational neuroscience
• 3D structure from motion

Background:

• The traditional view posits the visual system aims for veridical 3D reconstruction using optic flow and egomotion.
• This optimal model predicts accurate perception under natural viewing but attributes errors to poor sensory data in labs.

Purpose of the Study:

• To contrast an optimal visual system theory with a heuristic theory.
• To investigate whether the visual system derives 3D structure from optic flow velocity gradients alone, without egomotion or rigidity priors.

Main Methods:

• Two experiments were conducted using optic flow patterns generated by observer motion relative to two surfaces.
• Observers viewed these patterns both while moving and stationary.
• Perceptual judgments of 3D structure were recorded.

Main Results:

• Static observers perceived nonrigid structures from rigid surfaces, aligning with both optimal and heuristic models.
• Moving observers also perceived nonrigid structures, contradicting the optimal model's prediction of veridical rigid estimates.
• These biases in moving observers were consistent with a heuristic model relying solely on optic flow analysis.

Conclusions:

• The visual system's perception of 3D structure may be guided by heuristics based on optic flow velocity gradients, rather than solely by optimal reconstruction principles.
• Moving observers' perception deviates from veridicality, suggesting egomotion and rigidity cues are not always used optimally to derive 3D structure.
• A heuristic model, focusing on retinal flow field analysis, better explains observed perceptual biases in 3D structure perception during motion.