Monocular vision based navigation is an important issue in mobile robotics. Omnidirectional vision offers a large field of view and a complete information on the world surrounding the robot. We present in this talk an adaptation of parametric active contour segmentation to free space detection in omnidirectional images. We will discuss different classical methods and present some comparative results in indoor and out-door environments. The segmented omnidirectional free space will help the robot in real time navigation and environment exploration tasks.