From Video Matching to Video Grounding

This web page contains additional results (videos) for the paper with title From Video Matching to Video Grounding by Georgios D. Evangelidis, Ferran Diego and Radu Patrice Horaud, presented in 4th ICCV Workshop on Computer Vision in Vehicle Technology: From Earth to Mars (ICCV CVVT:E2M 2013).

Abstract


In this paper, we address a relatively new problem that we call video grounding. It essentially aims at reconstructing a video, as if it would be without foreground objects, i.e. cars or people. What differentiates video grounding from background estimation methods is that the camera follows unconstrained motion so that background undergoes ongoing changes. We build on video matching aspects since more videos contribute to the reconstruction. Without loss of generality, we investigate a very challenging case where videos are recorded by moving vehicles that follow the same road. Other than video synchronization and spatio-temporal alignment, we focus on the background reconstruction by exploiting inter- and intra-sequence similarities. In this context, we propose an MRF formulation that integrates the temporal coherence of videos and exploits the decisions of an SVM classifier about the backgroundness of regions in video frames. Experiments with real sequences recorded by moving cameras verify the potential of the video grounding algorithm.

Video Results


Highway Rural

Chen et al. [1] 

Chen et al. [1] 
 
Whyte et al. [2]
 
Whyte et al. [2]

Proposed algorithm

Proposed algorithm

 

References


[1] X. Chen, Y. Shen, and Y. H. Yang. Background estimation using graph cuts and inpainting. In Proceedings of Graphics Interface (GI) 2010 
[2] O. Whyte, J. Sivic, and A. Zisserman. Get out of my picture! internet-based inpainting. In Proceedings of the 20th British Machine Vision Conference, London, 2009