Background modeling via a supervised subspace learning

Farcas Diana (1) , Thierry Bouwmans (1)

1 MIA - Mathématiques, Image et Applications - EA 3165

Unsupervised subspace learning methods are widely used in background modeling to be robust to illuminaion changes. Their main advantage is that it doesn't need to label data during the training and running phase. Recently, White et al. have shown that a supervised approach can improved significantly the robustness in background modeling. Following this idea, we propose to model the background via a supervised subspace learning called Incremental Maximum Margin Criterion (IMMC). The proposed scheme enables to initialize robustly the background and to update incrementally the eigenvectors and eigenvalues. Experimental results made on the Wallflower datasets show the pertinence of the proposed approach.

Mots clés: Background Modeling Subspace Learning; IMMC

Domaines: Traitement des images [eess.IV]

Farcas Diana, Thierry Bouwmans. Background modeling via a supervised subspace learning. International Conference on Image, Video Processing and Computer Vision, Jul 2010, Orlando, United States. pp.1-7. ⟨hal-00536017⟩