Foreground Detection Scheme Using Patch Based Techniques for Heavily Cluttered Video

Abstract

Foreground objects can be extracted effectively by subtracting the background in the image frames, provided that an updated model of the background is available at any time. This is achieved by initialization (also called bootstrapping) of the background followed by its maintenance. In this paper, a patch-based technique for robust background initialization has been proposed, that overcomes the sleeping person problem. The proposed technique is able to cope with heavy clutter, i.e, foreground objects that stand still for a considerable portion of time. The method rests on sound principles in all its stages and only few, intelligible parameters are needed. Experimental results shows that the proposed algorithm is effective.