Multiple-Baseline Stereo Robust against Brightness Variation Depending on Viewpoints

DOI Web Site

Bibliographic Information

Other Title
  • 視点依存の輝度変動に頑健なマルチベースラインステレオ法
  • シテン イソン ノ キド ヘンドウ ニ ガンケン ナ マルチベースラインステレオホウ

Search this article

Abstract

Acquisition of three-dimensional information through images captured from different viewpoints is one of important themes to be solved in computer vision. Lots of stereo algorithms have been studied, and they have been applied to two-camera, trinocular, multiple-baseline stereos and moving-camera systems. Generally, the larger distance between cameras should be required for better depth accuracy. However, this often makes the correspondence issue more difficult. Brightness variation depending on viewpoints can be considered as one of the reasons in real scenes, which may be caused by non-Lambertien reflectance property of an observed object. In this paper, the authors propose a novel multiple-baseline stereo algorithm that is robust against brightness variation depending on viewpoints. It introduces a new cost function summarizing neighbor-camera comparisons and the authors prove its theoretical accuracy under the condition in which Lambertien assumption is comparable to a two-camera stereo algorithm that has the same baseline length. The experimental results by using real image sets that include brightness variation depending on viewpoints indicate it can show comparable or better accuracies not only to a two-camera stereo but also to a multiple-baseline stereo.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top