Evaluation of Background Subtraction Methods

In this paper we evaluate and compare six well-known foreground from background subtraction methods against a standard database. To be able to compare these algorithms objectively, we have chosen three challengeable scenarios from this database. The algorithms were applied to image sequences of length 100 to 800 frames. We examined the results thoroughly and supported the experimental results with a careful analytical study. We concluded that each algorithm might perform satisfactory in dealing with some particular challenges. That is, for example where we encounter background changes, illumination changes and flickering screens. This evaluation study assists the reader to be able to choose the most appropriate algorithm (among the six) for any particular situation.