Interactive planar reconstruction of objects and scenes

3D reconstruction from 2D images is an active research topic in the computer vision community. Classical algorithms like Structure From Motion (SFM) and Multi-View Stereo (MVS) are known to provide a robust framework for reconstruction, but fail to produce a dense plausible reconstruction. Some recent works have shown that making planar approximations of the scene help obtain more complete and visually pleasing results albeit, still approximate. However, many of these algorithms depend on image level cues like strong edges and lines which may be absent in textureless scenes. This has led to interactive 3D reconstruction algorithms. In this work, we present i3D, an interactive planar reconstruction algorithm. In our proposed system, we make planar approximations of the scene and with the help of user interactions, reconstruct the scene given as few as five to ten images from varied poses. As opposed to other more involved interactive algorithms which require the user to provide plane...

• Type: Others
• Year: 2009
• Where: Western New-York Image Processing workshop
• Authors: Adarsh Kowdle, Yao-Jen Chang, Tsuhan Chen
• Attachments: 1 file(s)
• Keywords: Computer Vision, Multi-image 3d Reconstruction