Experimental Results

We compared the proposed approach with classical statistical background models (SG, MOG , KDE), the reconstructive subspace learning models (PCA, ICA, INMF and IRT) and the discriminative one (IMMC). The experiments were conducted qualitatively and quantitavely on the Wallflower dataset.

Figure 1: From top to bottom: original image, ground truth, SG, MOG, KDE, PCA, INMF, IRT, IMMC with N=30, IMMC with N=100, IPCA-LDA. From left to right: MO (985), TD (1850), LS (1865), WT (247), C (251),B (2832), FA (449)