Figure 1

Principle of the background subtraction using independent component analysis. The input data is a combination of a reference signal, that is, background image (top left image), and a given signal, that is, an image from the sequence (bottom left image). ICA algorithm allows separating the input data into two independent signals: the one corresponds to the background model (top right image), the other corresponds to the estimated foreground without any details of the background (bottom right image).