Eric Lange
Well-known member
I think you make a really interesting comparison and point, especially with the audio.
There is enough of a safe range of on screen parallax to play with to make things very rich, comfortable and engaging. It’s just at the moment there are not very many good examples of folks that know how to “play well” within this range (I put this mainly down to the frantic nature of film production).
Here’s an audio analogy of what we have essentially solved (almost by accident) for s-3d.
So I have been lucky enough to be on stage with a full Orchestra. Here you hear the faintest sounds as well as the massive crescendos and massive bangs from the timpani etc. Take a Beethoven 9th, front row or closer. In this piece of music there are incredibly quite and delicate passages, and some massive belters with a chorale opened up to the max. The real dynamic range is massive, yet when we record we try to capture as much of this dynamic range as we can. However, when this is cut to lets say CD, there’s a massive amount of attenuation and sculpting of the whole sound envelope, to bring up the really quite passages and to bring down the full-on belters. Even then when you go play the CD, you will find that you will constantly have to turn up and trim down the volume as the dynamic range presented is still pretty anti-social if anyone else is around. I.e. if you set the volume to only one setting so that you could hear the quiet passages clearly, then when the loud finale kicks in you are still blasting heavy duty speakers by the end. Never the less this attenuated recording you still capture and perceive all of the fine high frequency elements that have a low amplitude…
What we have found is that you can effectively attenuate all of the macro parallaxes or large scale parallaxes that correspond to the macro structure of a scene without affecting the fine three dimensional high frequency 3d elements. What this means is that you can shoot with a wider IOD to capture this fine 3d detail, and then separately rein in too larger macroscopic parallaxes. The process that achieves this is very elegant but required some serious capability to be engineered first (for completely different reasons). So in other words we can re-render in a real-time fashion (once the data sets have been prepared correctly) the stereograms into safer ranges of viewing parallax without compressing this fine 3d detail. So basically you make the very best of the available safe ranges of parallax. Basically we had to take the “stereogram” up a couple of explicit dimensions to solve the problem but it also yielded a new type of real time data set called a Coherently Stereo textured Model that is 400x more efficient at being rendered over conventional CG data sets and looks completely life like as nothing from the original stereogram is lost.
We have been working in slightly different application areas and in essence have had this capability since late 2001 to early 2002. I had assumed that by now cinematic S-3d would be doing a pretty good job without our techniques by now. But it seems that these problems remain largely un-solved and its now 2011. The second issue is that unless people are able to understand or discover for themselves exactly what the problem is with s-3d, then they are not going to be able to understand what the solution to that problem really is either. At the moment these problems seem to be persistent and although our software is very capable it is not primarily designed or aimed at the stereo-cinematic market (such that it is), although it does handle very high res stereo sequences (and does a bunch of other stuff related to serious metrology). Next few months we put together some very high res, definitive data sets (we hope) to really prove the point. First we wanted to develop a safe area of application in our “home markets” that nobody can mess with before we start getting into the mix with more aggressive cinematic/media industrialist potential partners and competitors. Unfortunately it takes a bit of a “war chest” to swim with the sharks and it helps to be a certain size so that you don’t accidentally get swallowed up.
Increased computer horse power makes this all possible as well as good software design that optimizes the division of labor between what human beings are best at versus what computers are most adept at performing.
So I would basically say that as soon as s-3d technicians get fed up with their current “Kludges” as you [RAVI] say, then I think this will open the door to much more elegant solutions that will indeed offer a very rich and comfortable s-3d experience. It’s going to take a bit of time when S-3d specialists first reach an asymptote with what they can currently achieve. I think that will characterize s-3d development in general, that everybody gets up to certain level and get bored or frustrated with that, and then there is a new set of breakthroughs and then things plateau again, and then new things are brought to bear or come into existence once more. This is why I keep saying there is so much more that can be done, even within a safe range of on screen parallax.
Technology transfer is an art and timing is everything, if you are too ahead of the game then you have to spend way too much to bridge a gap that will instead be bridged automatically if one is patient.