摘要： 在场景深度信息未知的情况下，由多视点成像产生的焦散失真是无法消除的。针对该问题，利用多视点成像中光束斜交的特性，提取失真图像中结构化图形中包含的场景深度信息，实现对单幅多视点图像的三维几何结构重建。重构适用范围包括所有已知函数形式的曲线，尤其为直线、圆等特殊几何图形的三维重构提供了专门的数学方法。

关键词: 多视点图像, 失真, 深度信息, 三维重构

Abstract: Without knowing scene structure, caustic distortions due to multi-viewpoint imaging can not be removed. Based on the skewed light rays in multi-viewpoint imaging, this paper extracts depth information from structured features in distorted images, and performs 3D Euclidean reconstruction from a single multi-viewpoint image. The reconstructing principle includes all curves of known functional form, and provides specially designed mathematical approaches for features such as space lines and circles.

Key words: multi-viewpoint image, distortion, depth information, 3D reconstruction

中图分类号: 

• TP391.41