The object detection is first step for all automatic videosurveillance system. It is also one of the most interesting, well focused and well addressed but still challenging topic in computer vision. This paper introduced a new foreground-background segmentation method based on codebook. The main objective of our proposed method is to reduce the complexity of background modeling using codebook. For this, we proposed a background modeling based on superpixels. The number of superpixels is automatically update according to the size of the foreground objects detected. We use some frame-based metrics to evaluate the proposed method. Experimental results demonstrate the performance of our proposed approach.