Guide background model update with object tracking

Background subtraction has been widely used as the first step in various computer vision applications. It aims at segmentation of current frame in a video sequence into background (the relatively stationary scene) and foreground (hopefully including the objects of interest). Among the most difficult problems for background subtraction methods, we will focus on how to update the background model efficiently and accurately in this paper. A modified codebook based background subtraction method is presented with its model updating guided by an object tracking module. Experiments are carried out to compare the presented method with the classical MoG background model and the original codebook based background model.