Single-View Reconstruction using orthogonal line-pairs

摘要

Multi-planar buildings and man-made structures are characterized by a profusion of parallel and orthogonal lines. In this paper, using orthogonal line-pairs as the primary feature, we describe an automatic algorithm to recover 3D structure of a multi-planar scene from a single image. First, we show how the presence of such regular angles can be used for 2D rectification of an image of a plane to a fronto-parallel view. Next, by exploiting this ability to rectify scene planes, we propose an automatic Single-View Reconstruction (SVR) method, assuming there are enough orthogonal line-pairs available on each plane. This angle regularity is only imposed on physically intersecting line-pairs, making it a local constraint. Furthermore, we also describe a novel algorithm to automatically segment planes within a scene, and discover their extents and adjacency relationships, using only orthogonal line-pairs. Unlike earlier literature, our approach does not make restrictive assumptions about the orientation of the planes or the camera view, and works for both indoor and outdoor scenes. Results are shown on challenging images which would be difficult to reconstruct for existing automatic SVR algorithms.