Wyner-ZIV coding of multiview images with unsupervised learning of disparity and Gray code