EXPERIMENT: Use 3-D Rig For HDR?

[Joe G.]

Hi there. If anyone has a 3-D rig with a beam splitter, we have been pondering the options for HDR photography (in a different thread), and this one sounds like a possibility right now.

Instead of using the two cameras for stereo side by side, use lots of ND on one of them, and try different exposures in order to get a very high and very low light version of the same image.

I don't know if there is an easy software solution to blend the two in the best possible way without a lot of post work, but it should be doable.

Anyone care to give it a try??????

[Roberto Lequeux]

I have asked for this before. It would be very cool to see it happen. However you would lose some sensitivity, and obviously mobility. The new DSMC's will have more DR but it would be very cool to see this happen with them since they will be more sensitive and much much smaller. I know there is a limit to what you can get out of a super extended DR while keeping the result realistic, but it would be so sweet to manipulate skies (or snow, sand, or any highlights)... no need to replace them unless they look boring... though you might still need to do a lot of rotoscoping for full control over the sky itself.

[Noah Kadner]

But how would you optically converge the images? In order for HDR to work you have to have almost identical perspectives, otherwise it will blur anywhere the two images converge- which is of course the point of a 3D rig.

-Noah

[Frank Jonen]

@noah not necessarily blurry, you can converge the HDRs by aligning the edges after a 'find edges' pass. Then just cut into the image a little and roto your f-stops. Autodesk Toxic has a direct control for that but you can do the same in Nuke or Shake as well with simple rotoshapes and some tracking.

[Roberto Lequeux]

Also, doing this at 4k for 1080p, or 5k, 6k! You have so much detail that keeping sharpness high would be a piece of cake at 2k. Also, if you are splitting the beam it is a matter of making sure your optical paths are properly designed, calibrated, whatever... I don't know enough about optics, but all that is designed properly, and possible, then it shouldn't be an issue.

I wonder what would be the sensitivity of the "shadows camera" with half the light? That would dictate the exposure you would need to light for in order to avoid too much noise in that camera. Or you could just use the sensitivity of the "highlight camera" when you only need to get a normal look. Or just shoot with a single camera for night stuff which I suppose is perceived by the human eye to be lower in DR anyway. Right?

DSMC should be perfectly suited for this sort of application. With the right optical design for the splitter mumbo jumbo you might even end up with something equal in size to an F35! (?!)

I guess if your ratios aren't there you can't make it better with post, but you would always get all the info and you wouldn't have to fight to sandwich the light so much so you might save time.

[Tom Gleeson]

we have used a 3D rig to do this very test. Taking the interaxial distance to zero and aligning both cameras through a mirror rig we shot exactly the same shot but at different exposures. We crystal locked two A Minima S16 cameras and shot Vision 2 7217 with a three stop offset.

The spec TVC is still going through editorial but hopefully soon we will process the images but at first glance the dynamic range of the film stock is awesome and it remains to be seen but the overexposed camera may have captured the full dynamic range presented to it on our sunny day. I will report back our results when they are complete.

Tom Gleeson
DOP Sydney

[Joe G.]

"I will report back our results when they are complete."

Very cool.

[Roberto Lequeux]

Great. What is the dynamic range of that stock?

[Craig Parkes]

we have used a 3D rig to do this very test. Taking the interaxial distance to zero and aligning both cameras through a mirror rig we shot exactly the same shot but at different exposures. We crystal locked two A Minima S16 cameras and shot Vision 2 7217 with a three stop offset.

The spec TVC is still going through editorial but hopefully soon we will process the images but at first glance the dynamic range of the film stock is awesome and it remains to be seen but the overexposed camera may have captured the full dynamic range presented to it on our sunny day. I will report back our results when they are complete.

Tom Gleeson
DOP Sydney

I am sure this will be something everyone (especially Kodak) will be very interested in seeing the results from. If it offers one more tool in the arsenal that can't yet be achieved by digital imagers (not just increased DR, but HDR photography) then it will hopefully show that film has a few more tricks up it's sleeve.

Note: I'm a huge proponent of digital, but I also really LOVE it when people use a tool to new and interesting effect, due to evolution of not only complimentary technology - stereoscopic rigs in this instance - but also competing technology - digital still cameras undoubtedly increased the popularity of HDR photography - now motion film cameras may get the opportunity to take that interest and reinvent it themselves to create an entirely new look. That's real progress in action!

[Roberto Lequeux]

How is HDR limited to film? If anything you could go as far as automating the HDR merge in digital with a box and record the HDR files directly. You'd be able to grade on set, which you couldn't with film.

If you mean that one film camera could achieve the same as two digitals then I would assume that two 14-bit DSMCs would more than clear the DR of any one stock. And everyone knows about film's DR being greater than any digital to date.

[Craig Parkes]

HDR isn't limited to film, however at THIS particular juncture it's a better medium for this experiment.

In the future that may well change - but that was my point about progress. Progress is not only looking at new technology, but looking at new practices and thinking about how we can reuse older technologies to create new art and vision.

[Roberto Lequeux]

Ok, but I still don't understand what data would be gathered from testing film HDR that is more relevant when it costs so much more and takes so much more time.

Do you mean because most big budget stuff is still being shot on film? So for now it would still be more relevant to see the results with film? I know the benefits of shooting film over digital, I just don't see why it would be a better medium for something like this when the nature of HDR is in fact something that would make digital match film in terms of DR. I see it as much more relevant to see if two digital cameras could "catch up" with one film camera in this sense.

If I found it a possibility to shoot with a pair of DSMC brains with an optical mumbo-jumbo in front of them that only ended up making the whole apparatus about the size of a film camera, with no magazines!, then I would be thrilled about that. While shooting with two Arri 235's would no doubt give you a bigger and heavier rig. Also, you'd be burning twice as much film where recording twice as much digital data is a walk in the park in comparison to the cost of the stock and processing.

Plus, the sole possibility of merging both digital streams with a "box" if someone made one and have this box on the rig itself would mean that your DP and director would be able to work in a whole new way. I am no DP, and shooting with a nasty video tap is not what I want to be creating with when I can not extrapolate settings and stocks into the final look in my mind, especially not if there is the option to create seeing the real deal, and more so if it is actually more affordable to do it with digital and get something even better than the second generation of a camera line that already looks so amazing.

[Noah Kadner]

Most high-end DPs know how to light and filter scenes to fit their stock or electronic latitude as shot not try to rely on HDR to save shots with too much contrast. I'm a much bigger fan of NDs, Polarizers, big lights, flags and silks than carrying around two fully massive camera bodies and knowing every shot will have to be electronically re-aligned just to gain a few stops...

Noah

[KETCH ROSSi]

I love HDR imaging and getting more and more done, just never have in off time to do the edits once I return, hehe.

In fact Digital has so much more power in HDR then film ever had, and Digital continues to move forward with new designs and technologies allowing ever more for fantastic HDR imaging, currently I use the 1Ds III, even so for the purpose of HDR a faster shutter camera like the 10fps 1D III will be better, and you can shoot up to 7 consecutive frames at increments up to 3 stops.

With Red you can definitely do HDR, both in a Beam splitter "L" configuration set up or an S3D set up on a parallel axis, but you only get two exposures at the time, so why not just shoot with one camera and expose differently per every shot, since what you doing is using the MOtion camera to extract stills doesn't make sense to use a complicated 3D rig for such a purpose, at list not that I can see.

ciao

[Roberto Lequeux]

Of course a DP can manipulate the contrast of a scene, otherwise they wouldn't be a DP. What I am talking about is being able to give that DP something that lets them create without having to fight the light so much. Also, make it so that you can save time and money since you don't have to pull every light from the truck.

Ketch, I wasn't talking about stills. And like i said earlier, I am more exited about doing this with the future smaller cameras.

[KETCH ROSSi]

Ketch, I wasn't talking about stills. And like i said earlier, I am more exited about doing this with the future smaller cameras.

Understood Roberto, and yes I'm as exited as you are, or possibly more, about doing some serious S3D with the smaller bodies of Scarlet and Epic, even if Epic be a bit larger then Scarlet, the design will make Perfect for S3D configurations.

ciao

[Roberto Lequeux]

Must be nice. Have fun. I am not into 3d (is that what S3D is?) but using the same thing for HDR might be loads of fun... perhaps something a bit more conducive for a specific effect however.

For example, I have a large portion of a project where I would love to get crazy skies for... I could replace them and spend budget into actual sky replacement, or we could hope for enough day with decent skies and do something like this. Or close, since he probably shot several pictures... still, two gives you most of the bang.

http://www.scottkelby.com/blog/wp-content/uploads/2007/10/hdr.jpg (http://www.scottkelby.com/)
By Scott Kelby (http://www.scottkelby.com/)

Of course, then you are at the mercy of mother nature when it comes to the clouds. And while two Epics would be small they may not be as small as you'd wish unless someone designed the smallest possible beam splitter for them since I assume all out right now are behemoths. Or are they? Can one be designed so that they are side by side? Or just closer together?

How small do you think they could get Ketch? You should know.

[KETCH ROSSi]

Great shot Roberto, you did a fantastic work on the final HDR Tone mapping, I do not very much like the over toned mapping of some HDR's were they seems more surreal cartoonish drawings, I much prefer instead Photo realism, with over sutured Vivid colors in it, with the Tone Mapping done were the vision of a real world is presented by its life characteristics in color and depth.


S3D stands for "Stereoscopic 3D" just so to make a differentiation from 3D, generally understood as 3D Animatin, vs. S3D filming real actors in 3D. I don't like very much when it is achieved by the use of a Beam Splitter configured in an "L" shape which requires for a very large and almost impossible configuration to use in any Shoulder or hand held configuration, instead I prefer when done by configuring the two cameras in a Parallel axis configuration (Side by Side) this type of rig can get as small as the cameras them self, think it of as if it was a Dovetail side ways were the cameras are mounted on side by side and they can be moved closer or apart from each other, and a secondary Axis to allow the cameras to be positioned in a V shape with the lenses pointing in words, this off course must be achievable by smooth movement of the rotation of the axis preferably done by a Will, just like in the ARRI Head.

This is what I prefer as it gives a far more realistic and a much better 3D depth to the images in a very realistic way, dramatizing so much more your characters and your story, no objects flying out on you form the screen kind of thing.

So this type of configuration is extremely light, very versatile as far as mounting options and easy to transport, so Scarlet, and Epic will definitely be a game changer in this type of Rig, right now the best rig in use is that of Pace, Fusion 3D, but it is HUGE!! So Huge in fact that the Brain and lenses are the one mounted on the Parallel Axis while the Rolls are mounted further away from the Rig, so to allow for faster movement and ease of control.

The Stereoscopic 3D rigs such as the one from ET (limited to rental for now) or the one form P+S Technic, will get so much smaller with Scarlet and Epic.

ciao

[Roberto Lequeux]

Great shot Roberto, you did a fantastic work on the final HDR Tone mapping, I do not very much like the over toned mapping of some HDR's were they seems more surreal cartoonish drawings, I much prefer instead Photo realism, with over sutured Vivid colors in it, with the Tone Mapping done were the vision of a real world is presented by its life characteristics in color and depth.

Oh, crap... Click on the photograph please. I am so sorry. I apologize to Scott Kelby (http://www.scottkelby.com/) for being lazy and only making his photo a link to his website, where I copied it from because I thought it was such a great example for the point I was trying to make. I should have also put his name below it.

I agree though, once you lose realism you lose the ability to affect people in that certain serious way. Of course cartoons can also be stunning works of art.

I hope someone hurries up and makes a beam splitter for Epic. I would suppose companies like P+S Technic will jump on it, fully appreciating the impact, as you do.

Is there a benefit to shooting 3D with a beam splitter as opposed to using cameras side by side? Is it so that you can get narrower distances from eye to eye?

[KETCH ROSSi]

Oh, crap... Click on the photograph please. I am so sorry. I apologize to Scott Kelby (http://www.scottkelby.com/) for being lazy and only making his photo a link to his website, where I copied it from because I thought it was such a great example for the point I was trying to make. I should have also put his name below it.

I agree though, once you lose realism you lose the ability to affect people in that certain serious way. Of course cartoons can also be stunning works of art.

I hope someone hurries up and makes a beam splitter for Epic. I would suppose companies like P+S Technic will jump on it, fully appreciating the impact, as you do.

Is there a benefit to shooting 3D with a beam splitter as opposed to using cameras side by side? Is it so that you can get narrower distances from eye to eye?


Haha! Got it, I tough you tucked and create the post workflow on it.

Most of the current 3D rigs using a Beam Splitter such as the ET rig and the P+S T rig, I imagine would have no issues at all to be sued for the Scar;et and or Epic, but I still would use such a configuration for a very limited amount of shots were you could not achieve them using a Parallel Stereoscopic rig.

To me it just is a better way of transferring the Emotion, Drama, and true nature of how humans perceive motion in real world, just accentuated by the use of artistic digital manipulation and specific Plane of Focus to direct the viewers attention to what matters most in the story, and this is why Cameron chose the Parallel Stereo System vs. the Beam Splitter for AVATAR, even if some scenes were still captured with the implementation of a Beam Splitter rig.

To me the most important thing in a Stereo 3D rig remains the capacity to implement as much as possible the technology and make use of it to replicate as close as possible the natural eye movement and interpretation of how we Humans thru our eyes capture and process images, especially motion, as if not done right will never look right, no matter how much post you do, and crappy 3D is not bad, it will make you sick till you vomit, and if you have a strong stomach and you watch just the same, will put such a strain in your eyes that they could even get damaged.

The narrow distance that you look for to achieve in a Parallel Stereo rig is generally is set @ 2.5" (64mm) measuring from the center of each front lens element, from here they can be moved closer to each other or further apart to create e desired effect of Shrinking or enlarging the photographed subjects out of their original proportions, for any reasons of artistic reason in the Storytelling.

But what is more important is the capacity of Converging the point of view of the lenses when configured in such a Parallel Stereo rig, allowing for changing the Plane of Focus as needed to guide the viewers attention were the Director wishes to, as we normally do in 2D Photography while we use DOF to direct the viewers attention were we like.

[Steve Hertler]

The 3D rigs that Element are producing will be convertible between Beam Splitter and Parallel modes. This is because we realized very early on that in order to generate a proper 3D image for EVERY shooting situation, most productions will need to utilize both forms of image acquisition. The reality is that while a Parallel Rig can function properly in most circumstances, as the subject moves closer to the camera you require an interocular distance that can simply not be achieved without a Beam Splitter system. Conversely, as the subject moves further away and the focal length gets longer, a Parallel System becomes the most appropriate tool. This is not to say that great results can't be achieved by just using one system or the other...but ideally both systems would be employed dependent on subject distance and focal length.

Ketch said it right when he said that, for some stereographers, replicating the human experience will be the key. Personally speaking, the optical center of my eyes are about 2.5" apart from one another. In order to best replicate my experience of viewing an object that is about 5' from my face, a Beam Splitter would be the only way to do it "realistically".

EDIT: I thought about this some more, and decided I needed to be a little more specific. Theoretically, with Iconix cameras in parallel you could get an interocular distance of about 1.4"...although I don't know of anyone who is using an Iconix as an "A" Camera. With SI-2K Minis in parallel, you could get a minimum interocular of 2.5"...but with no travel left to move the camera's closer in the event that artistically this were preferable. With the RED-One the minimum interocular achievable with a Parallel system is about 6.5".

[KETCH ROSSi]

That is a fantastic news Steve, to hear that your S3D gear will be able to accommodated both style configurations of Parallel and Beam Splitter 3D rig, as I mentioned more then once I prefer very much the look and fill achieved by Parallel system as I fill it gives some what a more realistic point of view as the human eye would, in turn involving the viewer in a greater experience, but nonetheless a Beam Splitter configuration is essential when having shots in your story that requires the focusing of either your subject or scenery been closer then 6'.

So I truly look forward in knowing more about the rig, I do have a big hope for the release of the Scarlet FF35 and make a rig for them, then later next year for the Epic FF35 or 645.

I know at the moment you guys are only renting a number of units in order to provide great care to your customers on set, but I hope that the release for sale of the rigs for Scarlet will be available at the same time that Scarlet becomes available.

I have been pushing production back, and now getting ready for the first of many S3D films, so I hope that time will be on my side.


On the fact of what S3D really should be about, at list from my point of view, experience (very limited yet :) but passionate ) is exactly that, you must, both as a FilmMaker and a Stereographer provide the viewer with such an experience that it so real life, that the difference from standard movie watching must be HUGE!! It is in fact said that an S3D film experience should closely replicate true life, and that in this way the Brain it self should produce more NEURONS, during the viewing of good S3D films, and in turn such a stimulation will give just a great pleasurable movie going experience.


EDIT:

Steve Hertler

"Quote:
EDIT: I thought about this some more, and decided I needed to be a little more specific. Theoretically, with Iconix cameras in parallel you could get an interocular distance of about 1.4"...although I don't know of anyone who is using an Iconix as an "A" Camera. With SI-2K Minis in parallel, you could get a minimum interocular of 2.5"...but with no travel left to move the camera's closer in the event that artistically this were preferable. With the RED-One the minimum interocular achievable with a Parallel system is about 6.5". End Quote"


Correct Steve, and this is why for now I have decided to wait for the release of Scarlet, the specs and size plus design make it for a far better suited S3D system, then any current camera in the market, I don't consider the Iconix, neither the SI-2K, even so they have been used successfully in more then one project, including " My Bloody Valentine" were their footage was intercut with that of RED>

For the style and story lines I like to tell I must have the possibility of been able to achieve the Parallel Interocular distance of 64mm plus very tight Convergence to allow for specific looks, I just hope that the Designs for Scarlet and Epic, will take serious consideration on their impact to the 3D market, and if any changes apply , for them to apply with the needs for 3D in mind.

The other consideration which greatly impacts the Interocular distance possibility is also the almost forced use of Zooms for S3D shooting, as it just makes no sense in the use of primes when the time for setting up the rig is so very long, but hoe that gets better, for now, at list from what I have been able to experiment and learn, Zooms are just the way to go.


ciao

[Radoslav Karapetkov]

I guess that a two-camera HDR rig won't be necessary for the entire shoot, just for specific shots that need the latitude.

So, it's a viable option.

[Peter G]

that sounds like a great idea. using a beamsplitter for hdr-video.

i certainly will give it a try and keep you posted.

it so happens that i have a beam splitter stereo rig to spare at the end of december.

[Dan Hudgins]

I made a large number of posts on the subject of using two REDONE with a 3D rig here about a year or so ago, and put many drawings up.

There are also options to use relay optics so you can shoot with one prime lens to avoid issues with follow focus shots and or zoom.

My DI software can free resize the 4K images and rotate them to get alingment, it can put the right S curves on and fuse the two image sets for each shot, and it is freeish to try.

==

The only problems you might have are:

1) Is the SMPTE LTC lock within 1/10 frame of so

2) For the over-exposed image will the RED ONE sensor solorize and give the brightest highlights the "black sun" problem.

You can find the answer to #2 by overexposing your RED ONE several stops and see if you get any solorization, if not then if the sync is close you should be free to get more dynamic range for the shots you need it for.

As for alignment issues, projected 35mm film prints are close to 1280x720 on the screen, and the OLPF reduces the contrast of single pixels so you do not have a 4K image on the single pixel level in the color information, there is quite a bit of blur, if you try to sharpen the images you will get aliasing. So you should concider the so called 4K images an "oversample" of the required RGB resolution of 1280x720 for transfer to 35mm film to look as good as something shot on ECN or better as HDR.

==

If you want to see how the HDR images will look without going to the cost of getting a rig setup, just shoot a more or less static subject with your single REDONE, the process the two shots with exposure spread into a single HDR shot and see how the noise looks on playback. Costs nothing, if it looks good then go ahead and try two cameras, if not save your trouble. If anyone in the Bay Area wants some help doing tests I have the free time to help with such HDR tests and frame processing at 4K etc.

[paulherrin]

It should be obvious that for any HDR shot - you want two or more images that are perfectly identical, yet represent different exposures, or value ranges of that same composition. Since a stereo rig is purposed after the concept of two slightly different viewpoints of the same scene, you would want some kind of mirroring rig to capture the exact same image on two cameras - exposure bracketed (or ND filtered seperately).

I believe there are some other options as well - with a single camera. The type of shots that need HDR treatment, are (obviously), those that contain a high dynamic range of values - typically scenes with uncontrollable lighting, such as sweeping landscapes. When lighting is under direct control - I believe it's always best to light for the camera's native range. However, say a landscape shot with motion control (or still for that matter) could be bracketed out and shot two or more times in succession and toned in post. But this is a viable option only on rare occasion. That's why RAW is so great - it holds a greater latitude of exposure. In theory, you could bake out multiple exposure levels of the original RAW files and then composite for an HDR tone. Though you may lose some quality as opposed to other means (increased noise) - it may be made up by the overall achievement of the image. It should be noted that the RAW may still not carry the full dynamic range of the image in some or many situations with high ranges of light levels.

By manually adjusting one frame in photoshop and recording your actions, you should be able to make a simple script and run each frame through. That seems to me like the best way to tone it. All this can be practiced with a Digital SLR. Shoot a sequence RAW or exposure bracketed (or both!), bake it out. Run them through the toning - and be impressed with yourself. Then consider that the RED will (make you) look at least twice as good :)
EDIT: Evidently, Photomatix (http://www.hdrsoft.com/) (under $100) supports some batch processing that may be more easily done than trying to wrangle a bunch of files. A trial version of the standalone or plug-in software (complete with free watermark) can be downloaded to fully test this. As far as functionality, I'm honestly not sure how it compares to Photoshop (or other HDR processors for that matter). Experiment.

Check out this article (http://chadrichard.typepad.com/timelapse/2008/06/how-to-make-an.html) about capturing time-lapse HDR - this equally applies to traditional frames as it does to time-lapsed stills.

Those are just my thoughts. I'm no expert on it, I don't have much experience with HDR recording or stereoscopic rigging - and I admit there could be other viable options out there. If I was doing really serious work - I might check out two REDS with some kind of mirroring rig, especially exploring some way to use one lens for both cameras. However, we can imagine the expense (time) involved in managing this set up, so, I do think for many applications it would be quite reasonable all around to work with a single camera technique.

"Better than what they had."

paul herrin