Real-time segmentation of moving objects in a video sequence by a contrario detection

In this article we present a real time algorithm for detecting moving objects in a video sequence taken with a fixed camera. When a background estimation is known, the algorithm is able to detect moving objects and locates them approximately. The method is based on a comparison between each single image of the sequence and the background, this gives us a fast algorithm which can be applied on a video sequence with a rather high frame rate, but on the other hand, we are also able to work with sequences having a very low frame rate. We use an a contrario model, assuming that moving objects yield locally major changes in the topographic map. Experimental results show that the algorithm is very robust to noise and to the background estimation. Moreover, the algorithm needs few parameters, which are easy to estimate.