Utilizing the 3-meter horizontal baseline to resolve occlusions and synthesize sub-pixel detail in the primary view.
Parallax Warping: The secondary (Right Eye) view is warped into the perspective of the primary (Left Eye) using pre-computed homography and real-time disparity.
Neural Fusion: The kernel identifies areas where the secondary view has clearer "un-occluded" samples and blends them into the primary reconstruction, effectively tripling the usable pixel data for the CNN.