Objects detection by expectation-maximisation algorithm application to football images

In This paper we present a supervised method of image segmentation based on the statistic approach expressed in an hybrid space constituted by the three relevant chromatic level deduced by histogram analysis approach, this technique may the possibility of adapting the treatments to the local context of image with a little priori knowledge. This method has been applied on colour images issued from a soccer video of football sports. The obtained results show how his method reconstructs faithfully the size of different region while discriminating textures areas. We next study the influence of statistical parameters and chromatic level on these results.