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 :)

HDR Mode...

Bena said:
My guess is they are shooting 1 frame in Raw, then Processing it out to 2 or more exposures and then combining them for output. Just requires fast cpu and lots of math.
Click to expand...

The problem of combining many shorter exposures is that it also adds read noise. For example, if you took 10 exposures within 1/48 second and added them together to recover the shadows and averaged them together for the highlights frame, it would add some noise for the shadows because sensor read noise of each exposure would be added together. But, if the sensor is very-low noise like the MX is.... :sneaky2:
 
Noah Yuan-Vogel said:
Seems to me the motion artifacting could be significant. to significantly extend dr youd need samples that are quite different in exposure time, ... am i describing that correctly?).
Click to expand...

The one thing is because the frames start exposure at the same time it might be somewhat easier to re-align them to eliminate motion artifacts etc... this of course would mean that the final frame might be downscaled a bit (I'm just speculating here because I see and agree with the problem you mention).

On thing is if you had multiple frames of different densities you would not necessarily need all three for image detail info. You could use some of it for image detail and some for exposure density and color information only. (I don't know the actual image processing theory here, just speculating that to some degree you don't need to just "layer" the frame info, you can select what to use where and for what from each frame. You really need to get highlight and shadow info from the "alternate" frames the most.
 
The method suggested in my post #183 does not suffer from motion artefacting at all.
 
Well here is another theory :)

If the sensor can have different setting that enable it to see in darker part of the image, and it also have settings to see the brighter part AND you can switch settings fast enough. Then it is "just" a matter of capturing at both hi and low ends at the same time, and use whatever blending method you prefer. Maybe even every other pixel can have different settings, and use a clever de-bayer to process it.

..the fun begins when you try to shoot in 3D with one sensor AND in HDR ;)
 
redcode raw is the trick

redcode raw is the trick

Sorry, I'm a sucker for speculation. Expanding on the 2 brain theory:

The the secret is in Redcode RAW. Remember it is still a form a compression (Wavelet, if i recall correctly). With any compression, you lose information from the original sensor data.

The compression is applied to the raw sensor data. Assuming the sensor can actually capture say 17-20 stops of DR uncompressed. The Redcode RAW then compresses that down to less. Say 12-14 stops, throwing out the deep shadows and bright highlight ends. Due to limitations in the sensor, heat, and data rates, they simply cannot process the full DR of the sensor. (These are all assumptions, Graeme please feel free to kick me at Redday if this is totally off the wall).

Thus comes the 2nd brain. One brain compresses the low end of the true raw data, and the other takes the high end. Now you get the full range the sensor is capable of at twice the data rate, without worries of overheating. The end result is 2 r3d files of different exposure, merge them together with Redcine-X and output a new r3d with 20 stops!

Richard, I also liked your sampling the sensor idea. Would be cool it the digital sensors worked like the film layers theory you had.
 
I could be wrong, but couldn't the Arri Alev high/low gain path also be described as HDR?

http://www.arridigital.com/teaser/sensor

"The ALEV III sensor design provides 32 pairs of outputs. Each channel is divided into a high gain path (H) and a low gain path (L), resulting in 64 channels arriving at the 14 bit A/D converters. In the final images the lowlights are reconstructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful information in all 16 bits."
 
Woho, I have waited for this for years and years, finally some one starts to take advantage of what the digital technology may offer instead of trying to imitate analog film.

The video filming DSLRs was one step, 3D one other (RED seems to be prepared for it to) and this step is the next in the digital revolution.
I predicted this to come years ago and finally now here it comes and I’m not surprised it’s you guys delivering.

I don’t know but this might be the greatest and most important thing you have done yet, much more important than even the RED ONE if its delivers what it might! It’s a bit like the revolution started by the Viper and Genesis. 35mm is going down.

I hope the process really gives a HDR with at least three different exposures with at least 2 f-stops difference like the method “Indie Rentals” suggested. Of course there will be a motion blur issue in such case but hey, I’ll live with that minor problem every day.

Now I just have to wait for some other technologies that will revolve what most people think about a film camera and wouldn’t it be wonderful if you guys were the precursor for what to come to.

Like an included Z-channel and the possibility to change the depth of field and angle of view in post. Yes it will come, it’s just a matter of, I don’t know, I want it now.
http://graphics.stanford.edu/papers/lfcamera/

And of course, I would love to see a RED with a CCD or at least some sort of sensor without the rolling shutter and the skew and wobble problems coming with it.

Any way, this was really good news! Keep on going strong!
Regards /Pontus w (Sweden)
 
Leif Thomas said:
I could be wrong, but couldn't the Arri Alev high/low gain path also be described as HDR?

http://www.arridigital.com/teaser/sensor

"The ALEV III sensor design provides 32 pairs of outputs. Each channel is divided into a high gain path (H) and a low gain path (L), resulting in 64 channels arriving at the 14 bit A/D converters. In the final images the lowlights are reconstructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful information in all 16 bits."
Click to expand...

That's what I was thinking too. But the Arri sensor is doing a primitive tonemapping algorithm by the sound of it. As Jim stated the Mysterium X doesn't, which probably means that they're just combining multiple exposures into one floating point image with a higher dynamic range than the MX would be able to capture. Great possibilities there.
 
If this is true, then it is as big an achievement as the release of the original RED1 camera.

I remember the posts by David Mullen about having a 15-20 stops of initial range that can be adjusted in post to whatever is necessary artistically.

The wait is sure worth it. :yesnod:
 
Andrew said:
both my friend, both......
Click to expand...

I understand. What I meant was, on the Announcement, only HDR for stills was mentioned. Now, we know its both stills AND motion, so I was wondering if HDR MOTION was "something very interesting goes here" in the #21 spot. It would make sense. I find it to be very interesting...
 
The more I think of it, you could probably get quite an increase of range with just two exposures, and I still think the non-reset of the sensor off the read is the way to go. I can DEFINITELY see problems of ghosting under certain conditions (a fast pan) but there may be many usable ranges.

Of course, that brings up an interesting thought, if you could set some kind of threshold on the data in a frame where you ignore it, as if it is close to overexposed, or close to underexposed, and cause the blending to not just be the same across the whole frame but relative to the exposure level in the area. Just thinking out loud :)
 
Leif Thomas said:
I could be wrong, but couldn't the Arri Alev high/low gain path also be described as HDR?

http://www.arridigital.com/teaser/sensor

"The ALEV III sensor design provides 32 pairs of outputs. Each channel is divided into a high gain path (H) and a low gain path (L), resulting in 64 channels arriving at the 14 bit A/D converters. In the final images the lowlights are reconstructed from the high gain path and the highlights are reconstructed from the low gain path for an image containing meaningful information in all 16 bits."
Click to expand...

Yup. That can definitely be considered HDR. It will be very interesting to see how this compares with Red's HDR. Both in final image and in the techniques used to get there....

Also, the Arri method is the only one that I can picture the way it works in my brain. But if Red doesn't use gain or tone-mapping then I am stumped.
 
HDR doesn't seem to have a fixed definition anywhere, other than higher dynamic range than whats standard, which of course means when all this technology is standard, will it then still be HDR!? :)

With the ARRI sensor, I honestly don't get what the advantage is of what they are doing. The sensor is still reading once and they are passing the output of the sensor through two different gain stages post the sensor before the A/D. This means the noise on the high gain will be amplified much more than the low gain. Whereas, if the sensor is setup to get all the detail in a shadowy area, but takes a snapshot during that exposure much earlier for the areas that would get blown out, you are actually getting more detail out of the sensor.

So while you've solved some of the motion issues, I think you are going to add different amounts of noise to the the high and low paths which might not be pleasing. I don't see differing analog gain as a good thing.
 
Joel Kaye said:
That's the real question. We know the sensor gets what the sensor gets unless shutter or iris is adjusted. Adjusting IRIS causes DOF problems... ergo the shutter is going to get adjusted or it's some kind of on camera curve/shadow/highlight adjustment. The latter might be a workflow timesaver but wouldn't actually create an image that couldn't already be created in post right now.

If the shutter gets adjusted on the fly, captured and then baked into the raw data that would be pretty interesting.

Next question becomes - does that work on moving subjects? I don't currently understand how that could work.
Click to expand...

I don't think I ask stupid questions but no one ever seems to answer my questions around here anymore....

David
 
David Battistella said:
I don't think I ask stupid questions but no one ever seems to answer my questions around here anymore....

David
Click to expand...

I don't think it's that, I think it's the fact that almost everything here is speculation and your question regarding baked in or not would be total speculation as well. In my opinion, I doubt it would be baked in. Red seems to always take the path of most freedom in their decisions with regards to data so I'd say, they would be more likely to store all that extra dynamic range from the beginning and allow you to decide what else you wanted to do with it.

But of course, as I said, this is total speculation :)
 
Rick Burnett said:
I don't think it's that, I think it's the fact that almost everything here is speculation and your question regarding baked in or not would be total speculation as well. In my opinion, I doubt it would be baked in. Red seems to always take the path of most freedom in their decisions with regards to data so I'd say, they would be more likely to store all that extra dynamic range from the beginning and allow you to decide what else you wanted to do with it.

But of course, as I said, this is total speculation :)
Click to expand...

I would have to imagine it's baked in. If there were no side-effect then RED would be doing it for their 'normal' image mode. If there are side effects then some aspect of the process will be baked in.

So it's pure speculation but no matter the approach logically speaking I think we would have to conclude it's baked in.
 
You must log in or register to reply here.
Back
Top