This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Multi-camera Scene Reconstruction via Graph Cuts Vladimir Kolmogorov and Ramin Zabih Computer Science Department, Cornell University, Ithaca, NY 14853 email@example.com, firstname.lastname@example.org Abstract. We address the problem of computing the 3-dimensional shape of an arbitrary scene from a set of images taken at known view- points. Multi-camera scene reconstruction is a natural generalization of the stereo matching problem. However, it is much more dicult than stereo, primarily due to the diculty of reasoning about visibility. In this paper, we take an approach that has yielded excellent results for stereo, namely energy minimization via graph cuts. We first give an en- ergy minimization formulation of the multi-camera scene reconstruction problem. The energy that we minimize treats the input images symmet- rically, handles visibility properly, and imposes spatial smoothness while preserving discontinuities. As the energy function is NP-hard to minimize exactly, we give a graph cut algorithm that computes a local minimum in a strong sense. We handle all camera configurations where voxel col- oring can be used, which is a large and natural class. Experimental data demonstrates the effectiveness of our approach. 1 Introduction Reconstructing an objects 3-dimensional shape from a set of cameras is a classic vision problem. In the last few years, it has attracted a great deal of interest, partly due to a number of new applications both in vision and in graphics that require good reconstructions. While the problem can be viewed as a natural gen- eralization of stereo, it is considerably harder. The major reason for this is the diculty of reasoning about visibility. In stereo matching, most scene elements are visible from both cameras, and it is possible to obtain good results without addressing visibility constraints. In the more general scene reconstruction prob- lem, however, very few scene elements are visible from every camera, so the issue of visibility cannot be ignored. In this paper, we approach the scene reconstruction problem from the point of view of energy minimization. Energy minimization has several theoretical advantages, but has generally been viewed as too slow for early vision to be practical. Our approach is motivated by some recent work in early vision, where fast energy minimization algorithms have been developed based on graph cuts [6, 7, 12, 14, 20, 21]. These methods give strong experimental results in practice, as documented in two recent evaluations of stereo algorithms using real imagery with dense ground truth [22, 27]. The energy that we minimize has three important properties: it treats the input images symmetrically, it handles visibility properly, and it imposes spatial smoothness while also preserving discontinuities....
View Full Document
- Spring '08