Abstract. Vision systems are widely used in autonomous robots. Images are a rich source of information about the surroundings of robots and can be used to calculate distances of the objects in the scene to avoid collision, for tracking of objects and localization. In particular, omnidirectional vision systems have gained popularity and several different implementations for mobile robots. With those systems, a robot can have information of the environment in all directions with only one image. In a previous implementation an omnidirectional system with a hyperbolic mirror was used as a range sensor for a mobile robot navigation task. In that case, the robot needed to move to acquire a pair of images for the stereovision processing. To eliminate this movement, we present here a new mirror design, a two-lobe hyperbolic mirror to acquire one image with information suited for stereo processing as the same scene appears distinctly in each lobe of the mirror. This new system is presented with all the mathematical formulation required to the design of the two-lobe mirror. All these are explained in detail and the results of the stereovision processing algorithm that extracts range information from the omnidirectional image is shown for simulated images.