Self-Calibration for Metric 3D Reconstruction Using Homography

In this paper, our goal is to reconstruct metric 3D models for large structures (or other 3D objects) from uncalibrated images. The internal camera parameters are estimated by using homographic constraints. More specifically, we propose to use straight lines and vanishing points to compute the infinite homography. However, because vanishing points are very sensitive to noise, we propose further constraints on homography with multiple image views. After obtaining the internal camera parameters, the metric 3D models can then be accurately reconstructed. Some preliminary results are presented in this paper.

• Type: Conference
• Year: 2007
• Where: MVA
• Authors: Cheng-Yuan Tang, Yi-Leh Wu, Pei-Ching Hu, Hsien-Chang Lin, Wen-Chao Chen
• Keywords: Computer Vision, Internal Camera Parameters, Metric 3D Models, MVA 2007, Uncalibrated Images