1. Visualization of Scene Structure Uncertainty in a Multi-View Reconstruction Pipeline Shawn Recker 1, Mauricio Hess- Flores 1, Mark A. Duchaineau 2, and Kenneth I. Joy 1 1 University of California, Davis, USA, {strecker, mhessf, joy}@ucdavis.edu 2 Lawrence Livermore National Labs. duchaineau@cognigraph.com Vision, Modeling, and Visualization (VMV) Workshop 2012 Magdeburg, Germany 12 -14 November 2012 1

2. Multi-View Reconstruction 2 Bundle Adjustment ‘dinosaur’ dataset images from [1].

3. Structural Uncertainty Visualization 3

4. Volume Visualization Techniques Volume Rendering Contouring 4

5. Procedure … 5

6. Evaluated Test Cases Frame decimation simulation Feature matching inaccuracy Self calibration tests 6

7. Frame Decimation Graphs 7

8. Frame Decimation Results 30 cameras 15 cameras 10 cameras 8 cameras 4 cameras2 cameras 8

9. Feature Tracking Graphs 9

10. Feature Tracking Inaccuracy Results 10 0% Error 1% Error 2% Error 5% Error10% Error 20% Error

11. Self-Calibration Graphs 11

12. Self-Calibration Results 12 0%1% 2%5%10% 20% Principal Point Variation Focal Length Decrease

13. Conclusions and Future Work Presentation of a structural uncertainty visualization tool Continued visualization of computer vision Investigation of our cost function – Scene structure computation – Camera pose estimation 13

14. Acknowledgements This work was supported in part by Lawrence Livermore National Laboratory and the National Nuclear Security Agency through Contract No. DE-FG52-09NA29355 14

15. References [1] Oxford Visual Geometry Group, “Multi-view and Oxford Colleges building reconstruction,” August 2009. [2] V. Rodehorst, M. Heinrichs, and O. Hellwich, “Evaluation of relative pose estimation methods for multi-camera setups,” in International Archives of Photogrammetry and Remote Sensing (ISPRS ’08), (Beijing, China), pp. 135– 140, 2008. [3] D. Knoblauch, M. Hess-Flores, M. A. Duchaineau, and F. Kuester, “Factorization of correspondence and camera error for unconstrained dense correspondence applications,” in 5th International Symposium on Visual Computing, pp. 720–729, 2009. [4] T. Torsney-Weir, A. Saad, T. M´’oller, H.-C. Hege, B. Weber, and J.-M. Verbavatz, “Tuner: Principled parameter finding for image segmentation algorithms using visual response surface exploration,” IEEE Trans. On Visualization and Computer Graphics, vol. 17, no. 12, pp. 1892–1901, 2011. [5] A. Saad, T. M´’oller, and G. Hamarneh, “Probexplorer: Uncertainty guided exploration and editing of probabilistic medical image segmentation,” Computer Graphics Forum, vol. 29, no. 3, pp. 1113–1122, 2010. 15

16. Reprojection Error versus Angular Error 16 Reprojection Error Scalar Field Average Scalar Field