Embedding Fixation Constraints into Binocular Energy-based Models of Depth Perception