Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Its time to introduce a HDRI workflow

Take the 35mm adapter concept, now reflect the image to 3 HVXs (or RedOnes) instead of one. Each camera takes it's own exposures. If you could rig this up and synchronize the shutters, all you would need then is software.
 
victors said:
Take the 35mm adapter concept, now reflect the image to 3 HVXs (or RedOnes) instead of one. Each camera takes it's own exposures. If you could rig this up and synchronize the shutters, all you would need then is software.
Click to expand...

That would be possible stack three RED One's together, with some sort of optical splitter. The optics would be tricky though.
 
HD Bubaloo said:
Better explanation of the Forrest Gump shot your talking about.

You shoot the bottom portion of the frame of Forest running with a hard matte or plate blocking out the top portion of the frame. Then you shoot the top portion of the frame with the bottom hard matted out or with a plate. You take the two shots and optically (or now via DI) merge the shots into a final composite.

HDR does not do any blocking or plating of the master shot and its composite shot. It is essentially a locked down bracketed shot with one exposure setting or filter (usually an ND grad, or color grad) and sometime focus setting, and then you shoot another identical shot but with a different exposure setting or filter (usually without the filter), sometimes focus setting. After you take both photographs (which are framed identically, due to being locked down) you merge the two into a single composite by isolating the desired portions, exposure, filtration, or focus that you want to remain, into a final picture. This is why you cannot do this with video or any moving images.

Does that make better sense?
Click to expand...

Thanks for explaining. Hey! you're a New Englander. Sweet.
 
Aren't companies working on this now, but by stacking sensors on top of each other with different sensitivities?
 
was rolling in bed thinking about this and I thought,
Perhaps its possible to use one sensor and have other exposures be kept in buffer or recorded to on board cards simultaneously at the time of exposure and re combined as a single r3d file.
The burst rate is already there, it might be possible to have +2 -2 exposures, perhaps it might need more transfer speed to achieve this.
Relay lenses are already being explored.

I'm simply speculating here, the engineering genius is another.

let the dice roll.
 
Funny .. I was thinking about this earlier today .. and viola here's a thread about it.

I think it would require identical sets of optics and sensors, not just sensors being fed by splitters. You'd have to split the raw incoming light into multiple paths and into multiple cameras (synced cameras at that). Every time you split the light coming in you effectively cut it in half. This is the same problem we have in theatrical lighting with RGB color mixing systems.

Now if you could artificially (digitally) manipulate the RAW data from a single sensor to approximate different exposure levels in post, I don't see why that couldn't be done. But in still photography, they're physically taking 3-7 or more exposures and then merging them in Photoshop or another tool.
 
so why can't a very sensitive sensor (to compensate for light loss) just snap 3 bracketed exposures (shutter speed is comprised of 3 fast exposures combined into 1) per frame? the motion would appear too clean so you can apply some motion blur. or run the framerate 5 or 6 times faster, snap your brackets, recombine and playback at the same speed?

it doesn't matter how much motion is involved so long as you can snap the image fast enough. and if you snap enough fast images in a row at different exposures, you won't have issues recombining each bracketed image.

it's like slow motion on red. right now we get 120fps at 2k. that's 5x more than we need. if you play it back at 120fps, instead of 24fps, you get normal motion and extraneous frames that could be bracketed. it might not work for high speed action stuff yet but for normal motion stuff...all you need is the processing power to apply hdr to those frames in realtime.
 
IMO the only two realistic approaches are:

Delta Luminance. (Don't measure how bright something is after XYZ seconds, measure the first derivative or velocity so to speak of the exposure.)
Pros: Every pixel has almost infinite dynamic range. No ghost trail.
Cons: The brighter the pixel the faster the shutter

RdRGdgBdB Bayer pattern:
Pros: Global shutter possible. No temporal artifacts of any kind.
Cons: Quarter resolution. Aliasing and other Bayer artifacts.
 
I think what most of you are talking about has been addressed using OpenExr and it's support for 16 and 32 bit float per channel. You use a still camera to create the original HDRI image, then apply it to video using various post tools (Fusion, Nuke, AE with the right plug in). Most are using photoshop and raw image as part of the process.

The latest video cards from Nvidia and AMD/ATI support both 16bit float (aka half) and 32bit float.

Jump over to
http://openexr.org for better/more info.
 
I have explained this already. Until you can freeze time, it is not possible if we are talking about HDR in regards to moving pictures. In the stop-go animation world, sure. Anything else needs repeatable camera movement, operation, subjects, etc all in unison.

This is why Jesus invented PreViz and regular composites.
 
HD Bubaloo said:
I have explained this already. Until you can freeze time, it is not possible if we are talking about HDR in regards to moving pictures. In the stop-go animation world, sure. Anything else needs repeatable camera movement, operation, subjects, etc all in unison.

This is why Jesus invented PreViz and regular composites.
Click to expand...

This sounds accurate. Until we can consecutively capture the same exact frame at different exposures you don't have reliable frames to merge in an HDR process.
 
Very interesting thread. Didn't understand all (most) of it because I'm not really a technical wizard, but fascinating nevertheless.
 
I think a lot of people still do not fully understand HDR. If they did they would know it only applies to the still photography world, which includes stop-go animation.

Until you can suspend animation, HDR does not apply, but other compositing techniques do.

Because we have such excellent post solutions and FX software these days, and since almost anyone can afford After Effects, Color, and Photoshop, there really isn't a need for HDR for RED or any other motion acquisition device. As they have said for years, fix it in post. Or as I always say, shoot it the best you can to begin with, doctor in post if desired.
 
bry2k said:
I've been thinking about this since Red One was first announced. I think you can get about 90% of the way there with an HDR post workflow and Scarlett, since you're shooting RAW. You can (and I have done many times) obtain a very usable "faux" HDR by shooting RAW with a still camera and exporting over/under bracketed TIFFs, then combining them. The result is very similar to taking the same shot three times with different shutter speeds and even with still photography, it's sometimes the only way to pull it off if you're shooting a scene with moving objects. The same thing could easily (well, not easily, since it will be time consuming as hell) be done with RAW footage from Scarlett. Shoot your scene carefully exposed for the midtones, export +1/-1 or +2/-2 test frames to TIFF, dial in your HDR look, then setup a batch job to do every frame.

What do you guys use to do this in the still photo world? I've been using Photomatix Pro.
Click to expand...

:umm: ....

No. No. Noononononono.

This is like doing this mathematically:

(((36-8) + 36 + (28+16))/3)

----

HDRI isn't magic. It doesn't need multiple exposures. Unless you map it to real-world values like in a radiance file it's a plain old vanilla image.

If you had a sensor with *more than normal* dynamic range then you could capture a linear HDRI image without any processing at all. And if you didn't care about posterization you could even save it as an 8bit per channel image.

In 5-10 years we won't call the files HDRI we'll just call the files "files". Because for all intents and purposes they're exactly like any other picture. If we only had 4stop sensors we'd call 12bit RED HDRI. For that matter, I don't know that HDR even has any conventional definition of DR. I would guess the minimum would be 16 stops to correspond to half float, but it's not a term that has a definite meaning.

HDR is likd HD. It's not some super special state, it's a semantic expression of an image quality. We're human. We like to classify and categorize things. Even when there is no distinct difference between 1079p and 1080p. One is HD and one... is something undefined and weird.
 
IMTHATONEGUY: I see what your are saying but keeping true to the process, it is different. The same in the long run, but you cannot just re-define the existing process to fit a new ideology of what you would like the outcome to be.
Just give it a new name and don't point at an existing workflow and say its kind of the same when infact it has a process that is defined.

not busting balls, just being technically honest.
 
HD Bubaloo said:
It cannot work for video. You cannot do it via 3D because of the axis parameters. You would need to make a camera with an optical splitter and paths to 2 or more sensors. Then each sensor would need to be individually DSP'd with image controls for exposure, etc with independent outputs to separate storage devices. On top of that you would need to be able to apply digital filtration to portions of the each sensors frame, to get the desired outputs needed to take advantage of the final composites.

Better than inventing this new HDR camera would be to light your subjects properly, use excellent lenses and filtration, a nice matte box and Apple Color.

This works for still images because you can re-bracket the shot with different exposures, filtrations, etc to make a nice composite HDR in the end. Unless you master freezing time, you will not be able to re-bracket in motion pictures outside of virtual images.
Click to expand...

It can and it does.

Back in 2006 ARRI was working on double exposure HDRI capture.

You really only need 2 exposures to match the DR of 35mm film - which is the goal here - right?

I have subsequently designed the concept of capturing HDRI using optical prism split to two sensor, where one is slightly over exposed in relation to ones direct setting on camera, and the other is "underexposed" with the use of ND filter calibrated to match the required exposure. The second sensor gives You all the (normally clipped) highlights, while the first one gives all the details in the shadows. They are both in symmetrical optical path and the final signal is processed with on-board electronics and generates 16-bit images to be saved on a recording media in what-ever format that supports 16-bit.

It is actually very simple system and very much possible today...

The question is who will come out with it first...
 
If You wanna go overkill - You could use three sensors - one exposed according to your settings on the camera and the other two under/over exposed in relation to the central one. Mind You it is always the "overexposed" sensor that is getting the straight light, so the "normal" exposed sensor is already shielded behind ND and the "under" exposed is behind even stronger ND.

This way one can use the exact same sensors...

But two is sufficient to match the latitude of 35 mm film. Three will be great for EPIC...
 
You must log in or register to reply here.
Back
Top