Using Shading and a 3D Template to Reconstruct Complex Surface Deformations

Mathias Gallardo, Toby Collins and Adrien Bartoli

Abstract

The goal of Shape-from-Template (SfT) is to register and reconstruct the 3D shape of a deforming surface from a single image and a known deformable 3D template. Most SfT methods use only motion information and require well-textured surfaces which deform smoothly. Consequently they are unsuccessful for poorly-textured surfaces with complex deformations such as creases. We propose to combine shading and motion to handle these cases. There exist a few previous works which also exploit shading. However these do not provide an integrated solution, and they assume the surface and its deformations are globally smooth. Furthermore, most require a training phase and a priori photometric calibration for surface reflectance, camera response and/or light, which is often not possible. We propose an integrated solution without these shortcomings by jointly reconstructing the surface and performing photometric auto-calibration, considering an unknown light source which is constant and fixed in the camera coordinates. We evaluate with qualitative and quantitative results and show that it is possible to accurately register and reconstruct poorly-textured surfaces with complex deformations without any a priori photometric calibration.

Session

Posters 1

Files

PDF iconExtended Abstract (PDF, 2M)
PDF iconPaper (PDF, 8M)

DOI

10.5244/C.30.38
https://dx.doi.org/10.5244/C.30.38

Citation

Mathias Gallardo, Toby Collins and Adrien Bartoli. Using Shading and a 3D Template to Reconstruct Complex Surface Deformations. In Richard C. Wilson, Edwin R. Hancock and William A. P. Smith, editors, Proceedings of the British Machine Vision Conference (BMVC), pages 38.1-38.12. BMVA Press, September 2016.

Bibtex

        @inproceedings{BMVC2016_38,
        	title={Using Shading and a 3D Template to Reconstruct Complex Surface Deformations},
        	author={Mathias Gallardo, Toby Collins and Adrien Bartoli},
        	year={2016},
        	month={September},
        	pages={38.1-38.12},
        	articleno={38},
        	numpages={12},
        	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
        	publisher={BMVA Press},
        	editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
        	doi={10.5244/C.30.38},
        	isbn={1-901725-59-6},
        	url={https://dx.doi.org/10.5244/C.30.38}
        }