HDR in REDCODE RAW

jeffry alan said:
The method I am suggesting is a modern day "Technicolor", in that there are three IDENTICAL prints. One negative (RAW), three+ print exposures combined.

J*
Click to expand...


You can do multiple exposures for still work, because most of the time people are doing stills of non moving subjects.

I don't think we're there yet for HDR video, even though I saw some interesting research work.

That being said, work in floating point and cheat. There's nothing wrong with using available compositing and roto tools to replace elements that are over exposed or under exposed. It's done all the time, just that no one realizes it's an effect.
 
So to summarize:

-Whatever dynamic range you get out of any sensor is what you are stuck with.

-A simple curves editor is sufficient to reveal details throughout the dynamic range inherent in the RAW file.

-Multiple exposures using different shutter speeds is the only reliable method to increase the dynamic range when combined.

Is that basically correct?

J*
 
Actaully, for moving video you'd nee multiple NDs, not shutter speeds, as shutter effects appearance of movement. That would mean you'd need multiple sensors, some darker, some brighter, some inbetween, all taking the exact same picture at the exact same time for the same amount of time - not easy. Sure, there will be ways with optical flow simulations to ease up on some of the requirements, and interpolation techniques and the like, but....

You've still not solved the key problem with HDR imagery - how do you display it. Now some HDR stuff can look very good, but some, as we've seen, looks blinkin awful.

However, I'm sure RED will spark an R&D project for someone to develop new technique for HDR and amaze us all with the results. In the meantime, we'll just make the best RED camera we can.

Graeme
 
From my research online it looks like you can acheive an enormous dynamic range with multiple exposure HDR, but then you have to tone map it to monitors and such that have a limited dynamic range.
I guess that begs the need for supporting hardware to be engineered to ensure a continous workflow that maintains the quality of the images from aquisition to exhibition.

I hope that the Mysterium sensor at least will offer an extended dynamic range compared to the sensors out there today. And combined with an advanced image editor that can really tweak the shadow and highlight details, the output should convince companies to support the revolution.

J*
 
With multiple exposures, you can, effectively, have any dynamic range you're prepared to wait for :-)

As for the future, it's going to be interesting!

Graeme
 
Because you are still limited to a normal dynamic range (6-9 stops). The definition of (H)igh (D)ynamic (R)ange is that it expands that range significantly (12+ stops), and a simple curves editor would have to be expanded as well to reflect all of the increased tonal range.
Click to expand...

Right, I understand this. My point is with a single RAW file you only have so much dynamic range. It would be a whole lot easier to simply 'tone map' the entire thing down to an 8 bit image......
 
jeffry alan said:
So to summarize:

-Whatever dynamic range you get out of any sensor is what you are stuck with.

-A simple curves editor is sufficient to reveal details throughout the dynamic range inherent in the RAW file.

-Multiple exposures using different shutter speeds is the only reliable method to increase the dynamic range when combined.
Click to expand...
My first summary point.

J*
 
Hehe funny idea just crossed my mind. It's based on a high end photoprinter with the 8 color inks (dark and light versions of CMYK).

For the sake of easy math. Let's say a 4k bayer sensor is actually a 2k RGB sensor. Now if we could get an 16k sensor we could have a 6 color bayer pattern which by my probably completely wrong math would work out to a 4k RGB sensor. Just call it a Super Bayer. with say a 1/8 ND you could theoretically get a pretty significant dynamic range increase.

Optimally these wouldn't be square pixels, each sensor well would be two rectangular sub-pixels.

Now characterizing a chip like that might be a nightmare.
 
jeffry alan said:
From my research online it looks like you can acheive an enormous dynamic range with multiple exposure HDR, but then you have to tone map it to monitors and such that have a limited dynamic range.
I guess that begs the need for supporting hardware to be engineered to ensure a continous workflow that maintains the quality of the images from aquisition to exhibition.

I hope that the Mysterium sensor at least will offer an extended dynamic range compared to the sensors out there today. And combined with an advanced image editor that can really tweak the shadow and highlight details, the output should convince companies to support the revolution.

J*
Click to expand...


What are you trying to achieve?

The Red camera may have more dynamic range than any digital motion picture camera ever available.

How is that not enough?
 
Actually Thomas, I was trying to pry some information about the dyanmic range of the Mysterium from Stuart (So all could see) but he didn't bite.

I KNOW that HDR imaging is going to be improved by the Mysterium, I am just curious as to how much.

J*
 
im.thatoneguy said:
Hehe funny idea just crossed my mind. It's based on a high end photoprinter with the 8 color inks (dark and light versions of CMYK).

For the sake of easy math. Let's say a 4k bayer sensor is actually a 2k RGB sensor. Now if we could get an 16k sensor we could have a 6 color bayer pattern which by my probably completely wrong math would work out to a 4k RGB sensor. Just call it a Super Bayer. with say a 1/8 ND you could theoretically get a pretty significant dynamic range increase.

Optimally these wouldn't be square pixels, each sensor well would be two rectangular sub-pixels.

Now characterizing a chip like that might be a nightmare.
Click to expand...
Tu habla ingles?

J*
 
jeffry alan said:
Tu habla ingles?

J*
Click to expand...

correction: *hablas*

Summarization:

Bayer pattern = a sensor covered with pixels filtered to one of 3 colors:
Red, Green and Blue.

SBayer = a sensor covered with pixels filtered to one of 6 colors:
Red, 1/8thND Red, Blue, 1/8thND Blue, Green, 1/8th ND Green.

You would probably lose a lot of resolution, much much more than you would with a simple bayer pattern, but it would be like automatic bracketing. Effectively it would give an extra 3 stops of latitude, although I wouldn't expect all 3 to be useful.

If this hasn't already been done. Dibs on the patent.
 
Last edited:
Damn you Gavin!

I was just reading through this thread, and had just come up with pretty much exactly the same idea, just before I read your post on the subject.
 
It's already being done (not by us), Gavin. And yes, the interpolation to make it work is hell. We're not doing it. I don't think it would, today, help our overall image, but perhaps in the future some such scheme may be the way things go.

Graeme
 
I have maybe a dumb question...

Wasn't OpenExr created to address these issues easily in post? With the new graphics cards from Nvidia and soon from ATI, 16 and 32bit floating point per channel processing will be available to any software that wants to use it. If I understand OpenExr correctly you can add/preserve not only exposure settings but z point settings and lighting info without using multiple images (but you can if needed, since many 3d apps offer multipass redering).
 
Jeffry, that's a different question all together. Foveon is a very interesting technology, but it's not without it's own flaws, the major one being it doesn't run at 60fps at a decent resolution, making it unsuitable for high definition motion picture use.

Graeme
 
Graeme_Nattress said:
Jeffry, that's a different question all together. Foveon is a very interesting technology, but it's not without it's own flaws, the major one being it doesn't run at 60fps at a decent resolution, making it unsuitable for high definition motion picture use.
Graeme
Click to expand...
Sorry for the jump, I was thinking about the sensors.

I know that the concept of the Direct Image Sensor sounds good but there are a lot of issues still, and I didn't know about video application.

J*
 
You must log in or register to reply here.
Back
Top