Occlusion-Aware Video Registration for Highly Non-Rigid Objects Supplementary Material

Bertram Taetz; Gabriele Bleser-Taetz; Vladislav Golyanik; Didier Stricker

DFKI Kaiserslautern, DFKI Research Reports (RR), Vol. 1, 3/2016.


To give additional insight into our proposed method we present the optimization methods that we used to minimize the proposed energies of the occlsion-aware multi-frame optical flow (MFOF) as well as the global denoising of the occlusion probability maps. Furthermore, we added further examples, including dense NRSfM reconstructions, to demonstrate the capabilities of the proposed method.

occl-MFOF_Supp.pdf (pdf, 5 MB )

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz