What exactly is going on in the bottom right picture? It looks like we're taking some features from the second images, selected using the mask from the third, and applying them to the first image to produce the fourth. How does this process work? A reverse Google image search didn't turn up anything relevant, so I was unable to locate a paper based on that information alone.
How does Laplacian help to reconstruct meshes from point clouds (bottom left)? Also, what does the top-left image shows us?