Semi-Global Stereo Matching with Surface Orientation Priors

TL;DR

This paper introduces SGM-P, an extension of Semi-Global Matching that incorporates surface orientation priors to improve stereo matching accuracy on untextured, slanted surfaces with minimal additional computation.

Contribution

The paper proposes SGM-P, a novel extension of SGM that integrates surface orientation priors, enhancing performance on weakly-textured scenes without significant computational cost.

Findings

Surface orientation priors improve stereo matching on untextured surfaces.

Plane and Manhattan-world priors yield significant performance gains.

SGM-P adds minimal computational overhead.

Abstract

Semi-Global Matching (SGM) is a widely-used efficient stereo matching technique. It works well for textured scenes, but fails on untextured slanted surfaces due to its fronto-parallel smoothness assumption. To remedy this problem, we propose a simple extension, termed SGM-P, to utilize precomputed surface orientation priors. Such priors favor different surface slants in different 2D image regions or 3D scene regions and can be derived in various ways. In this paper we evaluate plane orientation priors derived from stereo matching at a coarser resolution and show that such priors can yield significant performance gains for difficult weakly-textured scenes. We also explore surface normal priors derived from Manhattan-world assumptions, and we analyze the potential performance gains using oracle priors derived from ground-truth data. SGM-P only adds a minor computational overhead to SGM and is an attractive alternative to more complex methods employing higher-order smoothness terms.