The neural representation of stereoscopic depth contrast

Area of Science:

• Neuroscience
• Computational Vision
• Visual Perception

Background:

• Stereoscopic vision exhibits contrast phenomena similar to brightness contrast.
• Lateral inhibition between disparity-tuned neurons is hypothesized to explain these phenomena.
• Existing models of lateral inhibition (e.g., retinal ganglion cells) involve linear subtractive effects, but disparity-tuned units introduce nonlinearities.

Purpose of the Study:

• To investigate the underlying mechanism of contrast phenomena in stereoscopic vision.
• To determine if a linear subtractive mechanism, analogous to classical center-surround, operates in stereo vision.
• To address apparent implausibilities arising from neuronal nonlinearities and sparse disparity information.

Main Methods:

• The study proposes a model where Laplacian-like center-surrounds are constructed from disparity-tuned cells.
• It suggests that interpolation can fill in missing disparity information between features.
• The model's predictions are compared quantitatively with experimental data on neighboring feature disparities.

Main Results:

• Evidence suggests a linear subtractive mechanism operates in the stereo realm, forming a stereo equivalent of the classical center-surround.
• Nonlinearities in disparity-tuned units can be overcome by constructing center-surround receptive fields.
• Interpolation effectively addresses the issue of sparse disparity information and matches experimental findings.

Conclusions:

• A linear subtractive mechanism, akin to a classical center-surround, is proposed to explain contrast phenomena in stereoscopic vision.
• The model successfully integrates neuronal nonlinearities and sparse disparity information through specific constructions and interpolation.
• The findings provide a quantitative explanation for observed dependencies on neighboring feature disparities in stereo vision.