In this paper, we present a topological algorithm which allows to fill contours images. The filling problem has been widely treated and it recently appeared that it can always be split into two different process : a generic topological process and a dedicated geometrical post-processing which depends on the application. Our algorithm, based on a 2D topological map description of the image, addresses the first step of processing. It is fast, generic and robust. Moreover, the complete topological description allows to easily integrate geometrical constrains and makes our approach an interesting basis for every filling process.