Individuation of 3D perceptual units from neurogeometry of binocular cells

TL;DR

This paper extends a neurogeometric model of stereo vision to develop a neural-based framework for local correspondence and global scene segmentation in 3D perception, emphasizing the importance of sub-Riemannian geometry.

Contribution

It introduces a novel neurogeometric approach integrating neural algorithms for stereo correspondence and scene segmentation using sub-Riemannian analysis.

Findings

Sub-Riemannian metric outperforms Riemannian distance in scene segmentation.

The model effectively organizes 3D perceptual units from binocular visual data.

The framework advances understanding of neural mechanisms in 3D perception.

Abstract

We model the functional architecture of the early stages of three-dimensional vision by extending the neurogeometric sub-Riemannian model for stereo-vision introduced in \cite{BCSZ23}. A new framework for correspondence is introduced that integrates a neural-based algorithm to achieve stereo correspondence locally while, simultaneously, organizing the corresponding points into global perceptual units. The result is an effective scene segmentation. We achieve this using harmonic analysis on the sub-Riemannian structure and show, in a comparison against Riemannian distance, that the sub-Riemannian metric is central to the solution.