We present a visual odometry system for short term self localization. It targets autonomous planetary exploration in an unknown rough environment. The singularity of the method is to estimate long elementary displacements at low frequency, 0.1Hz against more than 1Hz usually. Our algorithm is evaluated in a Marslike visual environment, on short and long trajectories, as a whole and step by step. We provide preliminary characteristics of its performances that can be used to design an autonomous locomotion system.