An efficient motion detection method based on estimation of initial motion field using local variance

Abstract

Background subtraction is a facile way to localize the moving object in video sequences which provide sufficient different intensities of foreground pixels from the background. It has been observed that the motion detection task can be more challenging in some complex scenes, which exhibit sudden or gradual illumination, varying speed of object, similarly colored background, and shadows. In that concern, we propose an efficient motion detection method based on the initialization of background, which is further used for the foreground detection. Under this formulation, we have adopted the local property of the initial motion field to get the suitable threshold condition to make it generic for the adequate visual representation of foreground pixel intensity. The effectiveness of this method can be seen when it is compared qualitatively and quantitatively to other well-known background subtraction methods. The experimental results show that it can work well in static and dynamic backgrounds condition.