Improvements on Vibe algorithm for detecting foreground objects

In order to improve the execution efficiency of Vibe algorithm in the process of background model updating, a time and space integrated method is proposed, which can not only take the advantage of previous memoryless updating strategy, but also improve the execution efficiency of the algorithm by combining both with-memory and memoryless methods. To remove the ghost in the initialization phase of Vibe algorithm, we present a method which combines Vibe algorithm and saliency map. For the Conservative update mechanism of Vibe algorithm, a foreground life mechanism based on the life cycle of foreground blobs is applied, so that wrongly updatings of big and slow objects can be avoided. The experiments show that the above three proposed algorithms give better performance compared with classical Vibe algorithm.