Box-based codebook model for real-time objects detection

Background modeling is the key technique of video surveillance. An improved box-based codebook modeling method is presented. The background model is constructed by encoding each pixel into a codebook consists of codewords based on a box model. It can handle cluttered background and allow moving objects in the training video sequences. The box-based codebook model representation is efficient in memory and speed compared with other background modeling techniques. The experiment results show the speed of building the codebook model and detecting objects over other methods.