Stereo matching is one of the most active research areas in computer vision and many algorithms have been developed to solve the problem of stereo correspondence. In this work, we propose a parametric model based on a new Bayesian formulation to solve the correspondence problem, using a doubly stochastic prior model that allows one to find optimal estimators by the minimization of a differentiable function. This approach also allows one to incorporate edge information to avoid erroneous matching in large regions with homogeneous intensities. Finally some experiments are presented, comparing the results with today's best-performing stereo algorithms.