Disparity map estimation under convex constraints using proximal algorithms

In this paper, we propose a new approach for estimating depth maps of stereo images which are prone to various types of noise. This method, based on a parallel proximal algorithm, gives a great flexibility in the choice of the constrained criterion to be minimized, thus allowing us to take into account different types of noise distributions. Our main objective is to present an iterative estimation method based on recent convex optimization algorithms and proximal tools. Results for several error measures demonstrate the effectiveness and robustness of the proposed method for disparity map estimation even in the presence of perturbations.