Exploiting Global Connectivity Constraints for Reconstruction of 3D Line Segment from Images, CVPR 2010

Exploiting Global Connectivity Constraints for Reconstruction of 3D Line Segment from Images

Arjun Jain Christian Kurz Thorsten Thormählen Hans-Peter Seidel

MPI Informatik

Abstract

Given a set of 2D images, we propose a novel approach for the reconstruction of straight 3D line segments that represent the underlying geometry of static 3D objects in the scene. Such an algorithm is especially useful for the automatic 3D reconstruction of man-made environments. The main contribution of our approach is the generation of an improved reconstruction by imposing global topological constraints given by connections between neighbouring lines. Additionally, our approach does not employ explicit line matching between views, thus making it more robust against image noise and partial occlusion. Furthermore, we suggest a technique to merge independent reconstructions, that are generated from different base images, which also helps to remove outliers. The proposed algorithm is evaluated on synthetic and real scenes by comparison with ground truth.

Results

A synthetic data set: Top row: example images from the input sequence. Bottom left: colour coded reconstruction error of our approach (blue indicates a low error, red a high error, and black an error larger than 0.5m). Bottom right: ground truth model rendered textured and in wireframe.

3D line reconstruction from a set of 20 photos taken with a consumer SLR camera in a low-light situation, which resulted in images with a high pixel noise. Images bellow shows the 3D line reconstruction of this scene. Many details are reconstructed including the tiles on the wall of the rightmost house.

Materials for Download

Here is the full paper in pdf format: [pdf]

The data sets made available here include the captured images, ground truth data, our reconstruction, camera calibration data, and the 2D detected lines. If you use the data, you should cite our paper. For any queries, please write to the first author. To browse the datasets, please click here.

Please note, that the camera parameters were estimated using the Voodoo Camera Tracker, and the datasets can be loaded and visualized using this free software.

Citation

Arjun Jain, Christian Kurz, Thorsten Thormählen, Hans-Peter Seidel
Exploiting Global Connectivity Constraints for Reconstruction of 3D Line Segment from Images
IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010)


  @inproceedings{JainCVPR2010,
     author    = {Arjun Jain and Christian Kurz and Thorsten Thorm\"{a}hlen and Hans-Peter Seidel}, 
     title     = {Exploiting Global Connectivity Constraints for Reconstruction of 3D Line Segment from Images},
     booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010)},
     year = {2010},
     month = {June},
     address = {San Francisco, CA}
  }