More HDRx...

[Sanjin Jukic]

7D in video is not any RAW so doesn't count here in this HDR discussion.

 

[Roberto Lequeux]

I disagree SJ, I think it doesn't belong in an HDR discussion because there is no way you can get 15 stops out of it.

Other than emulating an "HDR look" with controlled lighting.

 

[joshua csehak]

7D in video is not any RAW so doesn't count here in this HDR discussion.

Sry, went off on a total tangent. If only I could've went off on a tangent about my Scarlet's DR...

 

[Casey Green]

Jim, Jarred, Ted, Graeme, Deanan, and everyone on the RED team:

Haven't had time to be on here lately, but just wanted to say, it is simply stunning what you've achieved here. I have no more words at this time.

 

[Sanjin Jukic]

I disagree SJ, I think it doesn't belong in an HDR discussion because there is no way you can get 15 stops out of it.

Other than emulating an "HDR look" with controlled lighting.

Roberto,

HDR is not to get always 15, 16, + or more.

It's about to get maximum DR from every shot that would need it.

 

[Michael Olsen]

I think when we talk DR its all a bit subjective. Is DR what is shown right out of the camera (pre-grade)? Is it what can be "pushed or pulled" from that footage (post-grade)? It it what "most people" would consider usable? Here's one of the most popularly referenced 7D DR tests:

http://www.dvxuser.com/V6/showthread.php?187606-Canon-7D-meet-the-Stouffer-4110C-dynamic-range-test

Competent operator, competent test chart, competent testing methods. Result: 7 1/3 - 8 stops DR.

I haven't seen anyone able to shoot a chart that showed more than 8 stops of DR with the DSLRs in video mode. Then again, I stopped caring quite a while ago and just started shooting what I could with them.

 

[Bryce O]

I think when we talk DR its all a bit subjective. Is DR what is shown right out of the camera (pre-grade)? Is it what can be "pushed or pulled" from that footage (post-grade)? It it what "most people" would consider usable? Here's one of the most popularly referenced 7D DR tests:

http://www.dvxuser.com/V6/showthread.php?187606-Canon-7D-meet-the-Stouffer-4110C-dynamic-range-test

Competent operator, competent test chart, competent testing methods. Result: 7 1/3 - 8 stops DR.

I haven't seen anyone able to shoot a chart that showed more than 8 stops of DR with the DSLRs in video mode. Then again, I stopped caring quite a while ago and just started shooting what I could with them.

Not to disrespect Berry over there, but that test seems highly biased. It looks like they exposed the whites passed clipping just to get the least DR out of that Stouffer test. I have a much easier time believing a test like Zacuto's (11 stops in video mode), where they actually used a control variable for comparison (film stock) and people who know how to get the most out of the cameras (Bloom)... Instead of by a bunch of pissed off Panny heads who have a hard time believing that their HVX footage is old news now.

And yea, DR testing is VERY subjective. Just ask some photography guys... Nobody agrees on one test. You can blow out the whites and get something like 7-8 stops, or expose more for the blacks and see up to 11. It's all how you shoot in the real world.

 

[Gunleik Groven]

It's about to get maximum DR from every shot that would need it.

I think we disagree quite a significant bit as to how easily that is done...

We do also disagree somewhat on the definition as well, but that is another discussion...

My point is that you can do this:
http://reduser.net/forum/showpost.php?p=664190&postcount=507

within RCX with the right ISO/LUT and a curve + sat adjustement.
Try it... -

Anyhow.

This thread is going off marks, and I shamefully agree to have been contributing to that.

Discussing the 7D is not at all relevant for a number of reasons.

I very much want this thread to stay open.

Clean me out if that's what is neccesarry.

Sorry for any de-railing.

 

[Bryce O]

Anyhow.

This thread is going off marks, and I shamefully agree to have been contributing to that.

Discussing the 7D is not at all relevant for a number of reasons.

.

yea, sorry for any de-railing. My posts were mostly meant to bring up the issue of the subjectivity of DR testing, not just the 7D. It just happened to be the most relevant example. Which brings up a good point... Kudos for Red for actually posting real-world examples of HDRx and not just some stupid test chart showing some arbitrary marks and numbers. :thumbsup:

 

[joshua csehak]

I think when we talk DR its all a bit subjective. Is DR what is shown right out of the camera (pre-grade)? Is it what can be "pushed or pulled" from that footage (post-grade)? It it what "most people" would consider usable? Here's one of the most popularly referenced 7D DR tests:

http://www.dvxuser.com/V6/showthread.php?187606-Canon-7D-meet-the-Stouffer-4110C-dynamic-range-test

Competent operator, competent test chart, competent testing methods. Result: 7 1/3 - 8 stops DR.

I haven't seen anyone able to shoot a chart that showed more than 8 stops of DR with the DSLRs in video mode. Then again, I stopped caring quite a while ago and just started shooting what I could with them.

Well for one thing, he set his contrast to -2, when it should be at -4 (max). But all I'm saying is that in a real-world situation, I captured highlights spot-metered at f32, and shadows at f2.8, and f2 with some pushing in post. Take from that what you will. Curious what this extremely unscientific method will yield with the MX. Might have to stop by the rental house next week.

Again, total tangent, not relevant to the discussion of Red's freaking sweet HDRx™ technology; apologies.

yea, sorry for any de-railing. My posts were mostly meant to bring up the issue of the subjectivity of DR testing, not just the 7D. It just happened to be the most relevant example. Which brings up a good point... Kudos for Red for actually posting real-world examples of HDRx™ and not just some stupid test chart showing some arbitrary marks and numbers. :thumbsup:

+1

 

[Michael Olsen]

One more post to say: Sorry for the tangent. Back to how this HDRx business. :beer:

 

[Gavin Greenwalt]

Keep in mind that perceived color depth is both a function of its chroma resolution (bit depth) and the actual resolution of the image. A 5k 8 bit image will have greater perceived color precision than a 2k 8 bit image. Just like a monitor blends R G and B lights together to form a single perceived color with dithering if you have a black pixel and a white pixel right next to each other you can create perceived gray. So the Epic S35 sensor will have an advantage in that even at the same bit depth of sensor the increased spatial resolution can translate into an increased chroma resolution through dithering.

Personally, I like your definition better, but the beauty of the internet is you don't have to argue, you can just look things up: http://en.wikipedia.org/wiki/High_dynamic_range_imaging

Reliability of crowd sourced sources aside I don't find any contradiction to my definition in the Wikipedia article. HDR is a result not a process.

Note the citation for the wiki article:

"Images that store a depiction of the scene in a range of intensities commensurate with the scene are what we call HDR, or 'radiance maps.' On the other hand, we call images suitable for display with current display technology LDR."

The authors of the paper (and by extension rewritten Wiki Definition) assign no implicit demand that the source of an HDR image be multiple exposures or separate images. They also in the same definition clarify that LDR is only LDR with "Current" display technology leaving open the possibility of HDR displays in the future.

Another paper cited in the Wiki article:

Conventional SDR images (8-bits/channel) cannot acceptably reproduce HDR scnes, which is usual in outdoor conditions. (...) A straightforward solution to this problem is the use of HDR cpaturing devices instead of the conventional SDR ones. (...) (HDR cameras)'s increased cost has limited their use, while the majority of the existing vision systems are designed to use SDR cameras. Another possible solution is to aquire an HDR image by combining multiple SDR images with varying exposures.

In fact further along in the same Wiki article:

This method was developed to produce a high dynamic range image from a set of photographs taken with a range of exposures.

Another method would be... simply taking an HDR picture. Bracketed exposure is just one method to achieve the desired result: a "High Dynamic Range Image".

To quote ILM's OpenEXR page:

16-bit integer-based formats typically represent color component values from 0 ("black") to 1 ("white"), but don't account for over-range values (e.g., a chrome highlight) that can be captured by film negative or other HDR devices.

Note that they consider film an HDR source. (From a single scan.)

Lastly just think about it logically: if HDR meant "combined LDR images" then you could have a theoretical sensor with 32 stops of sensitivity be defined as SDR or LDR but a bracketed 18 stop image be HDR. That's silly. The single exposure has more dynamic range but can't be defined as HDR because of the method by which it captured the image? I think not.

 

[I Bloom]

I still feel like you're being too black and white. I think HDRx™ will stand the tests of time. But I also am relatively certain there are going to be times when MNMB will flat out kick the can and be unusable.

It's not an either/or situation.

Part of selling professional tools is to know your tool's limits and work around them. It can stand the test of time and have limitations to be conscious of.

Well, here's the thing, it's not exactly the same situation you are talking about. The difference is with HDRx™ it appears you are getting two images from the same exposure period, essentially two read passes for one reset pass. That means these two images are very special because there is a very direct relationship between them. I don't know of any other camera that does this.

Someone writing an algorithm to estimate motion blur on the highlight track, can do so by looking at the places where it doesn't match up well with the lowlight (longer exposure) track. That's a much better situation than writing an algorithm that works by tracking arbitrary images.

When I'm shooting I'm usually pretty wary of post based or digital solutions. An analog or in camera method, always seems to work in a more pleasing way. But Gavin I think you are jumping the gun just a little soon in your estimation of what is possible. At least get the thing in your hands and test it first

IBloom

 

[Sanjin Jukic]

Gunleik,

my suggestion is to get any good HDR app STANDALONE and try out for yourself with R3D files.

The probably you would be able to contribute to this discussion here about RX HDR out of R1.

Right now to be clear on this point I'm not discussing HDR out of Epic 5K just because this is another De-bayer thing too...

Also no any Voodoo but only a pure science...(precisely speaking mathematics)!!!

 

[Joseph Ward]

Red HDRx™ 3D Question?

Red HDRx™ 3D Question?

Quick question: Will HDRx™™™™™ work with 3D? If I shoot with two Epic Xs in 3D config, can I use HDRx™™™™™?

Good question, I was thinking the same. Red can you answer this?

 

[Gavin Greenwalt]

When I'm shooting I'm usually pretty wary of post based or digital solutions. An analog or in camera method, always seems to work in a more pleasing way. But Gavin I think you are jumping the gun just a little soon in your estimation of what is possible. At least get the thing in your hands and test it first

IBloom

I do have HDR motion blur free footage in my hands though. That's how a lot of CG passes come off the farm.

And I've spent many an hour painting in fixes by hand to the glitches produced by optical flow algorithms. (Some of which were created by The Foundry's line of products)

Even in the hypothetical situations where the optical flow track is 100% perfect. And it manages to create a pixel for pixel 100% accurate vector map--the blurring algorithms aren't perfect either. So even if RED happened upon alien technology with perfect optical flow, the vector blur created from the perfect track derived from multiple exposures would still be subject to glitches. Such is life. The world isn't perfect.

Easy example: Imagine a waterfall. Think about that for a second.

If you just blur the water then the rocks behind the water get blurred as well. If you blur the falling water to a large degree eventually it's just semi transparent but the rocks behind should be perfectly sharp and motion blur free.

Another easy example to illustrate would be a shot of a darkened store window with a puppy in a cage. You can see everything in the store window but a bright white truck drives behind the camera reflected in the window. If you blur the truck then you blur the puppy in the window. You can't just blur the truck in the reflection without blurring what's behind the reflection in the store. Well I guess theoretically it's possible but I've only ever seen that done in research papers--and even then it had its own host of glitches.

 

[Álex Montoya]

Do you know for sure that they are using optical flow algorithms to stitch both streams?

Bloom seems to believe that both streams are taken at the same time. Isn't that possible?

 

[Gavin Greenwalt]

Do you really know that they are using optical flow algorithms to stitch both streams?

Bloom seems to believe that both streams are taken at the same time. Isn't that possible?

Unless The Foundry invented a new method of deriving motion blur that's never shown up in a research paper out of the blue just in time for RED and never mentioned it in relation to their other VFX research projects, yes we can safely conclude they're using oFlow. An Optical Flow algorithm is what The Foundry won its Academy Award for and it's the basis of almost all of their plugins.

I would even go so far as to say it's a safe bet that it's not only based on The Foundry's oFlow, it's based specifically on their new GPU accellerated Blink code which is in the new Kronos.
http://features.cgsociety.org/story_custom.php?story_id=5776

The HPC team at The Foundry has been working hard on speeding up the core motion estimation engines that drive the Furnace range of plugins. The results will be seen in the new plug-in for After Effects: Kronos 5.0. This will be powered by 'Blink', the new, go-faster technology the HPC team have developed, to provide realtime retiming. Kronos relies on estimating motion in the source footage and then generating in-between frames, a process which is currently running at least four to five times faster when GPU-enabled. The HPC team at The Foundry is confident that by the time these plug-ins are available to buy, they will be even faster.

Both streams at the same time is possible and it does appear to be what RED is doing. But it doesn't solve the intrinsic problems with post production blur that aren't unique to RED or Optical Flow or even motion blur. The only system that would avoid these pitfalls is a camera which shot highspeed (which also might be possible for RED in the future). But it would take more than 2x the datarate to achieve (think 10-100x the datarate of a single frame) but might be the basis of their stills HDR solution. (Shoot a burst of motion blur free frames and then recombine them like multi-mass motion blur).

 

[Michael McVicar]

Quick question: Will HDRx™ work with 3D? If I shoot with two Epic Xs in 3D config, can I use HDRx™?

I know that there is so much to test and do in the brave new world of HDRx™ and Magic Motion™, but I was also was wondering about 3D. Genlock on 4 data tracks with quadruple the data rate is some wild stuff. Temporal adjustments in post, motion blur with highlights perception in S3D, lots of uncharted waters. HDRx™ could be a huge help in lighting for deep focus. Just the tip of the iceberg here. Exciting times!

 

[Sanjin Jukic]

Quick question: Will HDRx™™ work with 3D? If I shoot with two Epic Xs in 3D config, can I use HDRx™™?

Tom, HDRx it could be definitely 3D thing just because will work mostly with a sort of surreal or fantasy images...