Originally Posted by tomtastic
There hasn't been many newly released 2D to 3D conversions lately, is this because converters work so well or do we need more 2D to 3D conversions released?
Edit: Old movies, like Titanic and Predator, just to be clear, not new releases. Those should be native 3D anyway.
It is because it is very expensive to do a high quality conversion, and people who bankroll such ventures are uncertain there will be a profit. The public do not appear to be as thirsty for 3D as they used to be.
It is not because converters work well. Built-in converters in Blu-ray players or 3D TVs are:
(a) of interest to try out for novelty value, and
(b) some people (a small percentage) have reported using such converters routinely and enjoying the result.
If converters worked tremendously well, we would expect large sections of the population insisting on putting on their 3D glasses to watch their 2D television programs in simulated 3D. There is no evidence of that. I for one am very keen on stereoscopic 3D but the automated conversions I have seen have left me cold.
I have tried the built-in 3D converter in my 2013 model 65" Sony 4k passive 3D display. It gives each scene a 3D effect, is able to infer the correct order of layering of objects a lot of the time, but unfortunately does not transition well from one scene to the next, e.g. with a series of advertisements each new ad may be associated with an abrupt, disturbing, change in the 3D effect.
Reflections are very challenging even with painstaking manual conversion. With automatic conversion, expectations should be low. As a test, I have tried out 3D stills and video of a swimming pool with reflected trees appearing on the surface of the water and some small black marks appearing at the bottom of the pool. My Sony set is incapable of separating these features in 2D to 3D conversion mode. But in proper 3D mode you can see the reflections, the surface of the water, and the black spots at the bottom of the pool, quite easily.
Automatic conversions tend to fall down when the subject is very close to the camera position such that a true stereo view would provide quite different images for the two eyes. Imagine if you will a human face facing slightly away from a 2D camera. If the camera is placed directly in front of the person at about a metre away it will reveal one ear to a large extent, and the other to a limited extent, or not at all. It will reveal one side of the nose in full and the other side of the nose only partially. A detailed manual conversion, as was done with Titanic
(1997), will painstakingly insert the missing ear detail, and the missing nose detail, so as to arrive at a close approximation of the views that a 3D pair of lenses placed directly in front of the human face would have captured. (Or that indeed a human being would have seen, using left and right eyes.)
The result of the detailed manual conversion is that the person really does looks very close to the viewer as in real life, rather than as a magnified photograph, or magnified 2D video.
The conversion for Jurassic Park
(1993) uses a fairly exaggerated, but effective, 3D a lot of the time. However with closeups I don't perceive the special "spice" that true stereoscopic 3D gives in closeups, and that very detailed manual conversions such as done for Titanic
, come close to providing. In my opinion, it was a professional and effective, but not exceptionally well done, conversion.
A conversion that was particularly well done in my opinion was Harry Potter and the Deathly Hallows Part 2
. Very early in the movie we see the interior of a cottage near the sea (Shell Cottage) where Griphook the Goblin plots with Harry to gain access to a vault in Gringotts Bank. The close-ups of the human characters in the cottage are very realistic, indicating considerable attention to detail in the post conversion.
It will be many years before computers are able to achieve that high standard of conversion automatically. (The computers will likely require a sophistication of analysis akin to artificial intelligence and would need considerable processing time and more than one pass through the video frames to obtain video data from later frames that could be used to help manufacture missing detail for earlier frames). Currently such conversions are done with the aid of specialised software and take thousands of man hours per movie. Much of the work requires frame by frame human intervention.
When it comes to chaotic image content such as raindrops falling immediately in front of the camera position, or multiple layers of image content too difficult to separate and unravel from the 2D source (e.g. a school of 500 small tropical fish swimming close to each other) a realistic manual conversion may become unwieldy or impossible, and approximations need to be made to make the conversion task manageable. This could include blurring parts of the frame.
If a real time automated 2D to 3D converter is preferred because it can provide a milder 3D effect than some native stereoscopic movies provide, that is an understandable preference. However the better produced native 3D movies do not contain exaggerated 3D, for my eyes, anyway. I would cite as examples: Avatar
(except in the opening scene where Sam Worthington's character comes out of cryogenic suspension), Hugo
, Life of Pi
and the two Hobbit