Have been testing algorithms for reconstructing the 3D structure of a shot. This is part of an ongoing research project into digital processing of film in order to unlock aspects of a photographic image that are not necessarily visible in any given frame. It follows a kind of a surrealist position with respect to the photographic image - that the image has a kind of unconscious aspect.
Anyway, as a test I applied the process to a shot from the opening sequence in the Wizard of Oz. The algorithm cross references each and every pixel in each and every frame in order to reconstruct a 3D model of the shot. Once this is done it then becomes possible to "rephotograph" the shot in a different way. Here I'm using a camera position the same as the original, but with a shallow depth of field. The process is entirely automated. Automatism is a big thing in surrealism. In other words the result is not a photoshop job, ie. not done by eye. The algorithm took 48 hours to process the 283 frames of the shot. I thought it would never end at one point, but held out. It eventually completed.
I am going to state up front that I have personal misgivings of altering original works. But this is exceptional investigative work. I am staring at this image transfixed. I initially triggered on your reference to 3D and my anticipation was some sort of hybrid stereoscopic image. Instead you have transformed huge expansive DOF into crushed DOF rendering Dorothy as THE focal point.
Your algorithm is similar to somehow attaching a classic bellows system to the incredibly heavy Technicolor DF-24 Beam Splitter Motion Picture Camera used to film the wonderful Wizard of Oz, i.e.
