In this paper, we propose a method, which adopts a global optimization algorithm based on a hybrid scheme. It consists of a genetic algorithm (GA) combined with a local search process. In the GA phase, an efficient fitness evaluation procedure based on a non-linear camera model is proposed. The local search step is a non-linear local approach using the Levenbergh-Marquardt algorithm. The experimental results show that the proposed method performs very well in terms of robustness to outliers and quality of camera parameters estimation.