PDA

View Full Version : Low Dynamic Range (LDR?) Feature



Pawel Achtel
11-29-2008, 06:15 PM
OK, so most of the time, when we film outdoors we face problems associated with high contrast. To overcome those challenges, we wish we had more dynamic range. But, what if it is dull, low contrast scene and we want some punch in it? Do you sometimes wish you had more data in the mid tones so you could crush it for a more contrasty look?

I think there is a need for Low Dynamic Range for low contrast scenes (high sampling recording within low range of brightness values) just as much as there is a need for High Dynamic Range for contrasty scenes. I don't think one range will please all and having high range and high sampling is, clearly wasteful.

Would it be possible to set DR for the sensor output range for the 12-bit or, in case of Monstro, 16-bit quantization within just that range?

Just a thought.

Tim Hole
11-29-2008, 06:19 PM
I think the whole of the UK needs a punch for at least 11 months of the year. Grey is the word!

Mark K.
11-29-2008, 06:54 PM
Pawel, I would have thought that with all the data there (i.e. high DR capturing detail in both the highlights and the shadows) it would then be far easier to adjust your contrast in post, because you'd have all of the information there in the first place to modify (rather than having clipped sections of your footage that you can't modify at all).

Kyle Presley
11-29-2008, 07:14 PM
Pawel, are you being serious?

jbeale
11-29-2008, 07:40 PM
Is posterization a real-world problem with the current cameras? Having a very drab, grey scene means that you can set your exposure relatively high without clipping any peaks, eg. rating for ISO 200 or even lower. By setting your exposure correctly, you have mapped your available bits optimally to the scene contrast; you don't need to do anything else.

This presumes you still want a linear transfer curve and you want to maintain your black level. I suppose you could also allow raising the black level to something grey, if you're really shooting in a complete fog. That would allow even more contrast stretching, in theory. The problem is that while your steps are smaller, you haven't improved the system noise level, so I believe the end result would look much the same.

Pawel Achtel
11-29-2008, 09:06 PM
Is posterization a real-world problem with the current cameras?

Yes, this was a real and quite severe problem I experienced with HDCAM camera :bleh:. A very low contrast scene was stretched in DSP (modified gamma curve) in the camera. The result was posterization of the background and it was really bad because it was flickering like hell as the scene was changing very slightly. Simply there were not enough bits to represent mid-tones for the stretch.

Whilst you wouldn't see it immediately on a still image, the motion picture was total disaster. In such situation I would love to be able to map only about 5-6 stops of the dynamic range into 12 bits of data allowing smooth and subtle tones to be reproduced accurately.

After all, if your scene only has 4-5 stops of dynamic range, why would shouldn't you be able to use full quantization capability of your camera to record it? It makes perfect sense to me.

Pawel Achtel
11-29-2008, 09:24 PM
Pawel, I would have thought that with all the data there (i.e. high DR capturing detail in both the highlights and the shadows) it would then be far easier to adjust your contrast in post, because you'd have all of the information there in the first place to modify (rather than having clipped sections of your footage that you can't modify at all).

Grug, You would not clip highlights or need shadows because in low contrast scene they are not there. So, effectively you are wasting your quantization bits for no good reason, where as you only use fraction of the quantization range to map the existing brightness range. Then, if you want to stretch that range to achieve higher contrast image (from a low contrast scene), you find posterization does not allow smooth gradiends to be reproduced.

I Bloom
11-30-2008, 12:03 AM
Grug, You would not clip highlights or need shadows because in low contrast scene they are not there. So, effectively you are wasting your quantization bits for no good reason, where as you only use fraction of the quantization range to map the existing brightness range. Then, if you want to stretch that range to achieve higher contrast image (from a low contrast scene), you find posterization does not allow smooth gradiends to be reproduced.

Pawel,

Are you actually experiencing posterization in your attempts to bring out the contrast in overcast scenes on RED. I've shot many such scenes and I've never had any problem making them pop. Can you show us an example??

Plus with the Monstro chip you get 16 bits instead of twelve. So again. Do you really have an issue.

I'm going to go out on a limb and call shenanigans. I think this thread is obsurd.

Sorry,

I.Bloom

Pawel Achtel
11-30-2008, 12:56 AM
Pawel,

Are you actually experiencing posterization in your attempts to bring out the contrast in overcast scenes on RED. I've shot many such scenes and I've never had any problem making them pop. Can you show us an example??

Plus with the Monstro chip you get 16 bits instead of twelve. So again. Do you really have an issue.

I'm going to go out on a limb and call shenanigans. I think this thread is obsurd.

Sorry,

I.Bloom

Ian,

I haven't experienced it with Red One and I haven't tried. (I only had an opportunity to use my Red One during single day). But, I did experience the problem with HDCAM quite often, actually. This was underwater, in less than optimum visibility.

I do not have an example handy (it would have to be motion picture), but to ilustrate the conditions that I'm talking about I attach a picture taken with a still picture camera with neutral settings. It is not manipulated. This is how it looked. I can't see more than 5 stops of dynamic range in the actual scene.

On the very same dive in exactly the same conditions I took footage with heavily modified gamma curves and matrix on my HDCAM. Frame grab is attached. You can see why I adjusted curves heavily. Those are the same settings that, in certain conditions, cause posterization flicker on graduated colour (such as blue water).

However, I do agree that 16-bit quantization is extremely unlikely to cause such issues. The problem is I do not have 16-bit camera...do you?

J. Bernard Vallon
11-30-2008, 10:45 AM
Is posterization a real-world problem with the current cameras? Having a very drab, grey scene means that you can set your exposure relatively high without clipping any peaks, eg. rating for ISO 200 or even lower. By setting your exposure correctly, you have mapped your available bits optimally to the scene contrast; you don't need to do anything else.

This presumes you still want a linear transfer curve and you want to maintain your black level. I suppose you could also allow raising the black level to something grey, if you're really shooting in a complete fog. That would allow even more contrast stretching, in theory. The problem is that while your steps are smaller, you haven't improved the system noise level, so I believe the end result would look much the same.

Pawel, jbeale is right, this is exactly what you should do. By pushing your exposure to the right, you're capturing it in a section of your chip that the AtoD converter gives more color resolution to. In the very last stop of exposure all the way to the right of the histogram, there are 2048 levels in the RAW to describe the colors, which is the same amount as the ENTIRE REST OF THE HISTOGRAM. In a linear signal, every successive stop has the same amount of data as the entire previous exposure combined.

If you only need 5 stops, rate your camera at 200, or 160, open your aperture, and look at all the pretty colors.

Graeme Nattress
11-30-2008, 01:03 PM
Expose to the right will give the best results here. When the DR of the scene is < DR of camera, ETTR puts the light into the code values in the data which have the least noise. As long as you're careful not to clip, you're getting the best results available.

Graeme

Tyler Black
11-30-2008, 06:52 PM
Wasn't there a thread somewhere that pointed to a link to an article by Thomas or John Knoll that explained this? I can't find it now.

JD Holloway
11-30-2008, 07:36 PM
Given Pawel's shooting environment, I would guess he doesn't have the ability to rate the camera at 100-200 asa (read... expose to the right properly). Underwater shooting is always hurting for light, let alone colour correct light (where did red channel go?). Wide open all the time. I don't mean to speak for him, just guessing. Pushing the histos right would work... but it assumes you have the lumens to do it. It only works for certain environs.
J.

Justin Anderson
12-17-2008, 11:02 AM
I don't think the "expose to the right" rule applies to HDCAM though. Only RED. :/

Does it?

Lauri Kettunen
12-17-2008, 11:56 AM
Does it?

Well, say one is shooting in low contrast conditions and has a 8-bit AD converter in the camera, but the editing system has higher bit depth. In this case I guess it makes sense to shoot to the right, that is, set the exposure just under clipping (or a certain point of the gamma curve), and then bring it all back to the correct level in edit. As a result one may end up having more data/bits on the mid-gray level. (The rationale of this depends very much on the gamma-curve employed.)

Brian Langeman
12-17-2008, 12:23 PM
I don't think the "expose to the right" rule applies to HDCAM though. Only RED. :/

Does it?

Exposing to the right applies to every camera that records digitally. It uses up more bits no matter which camera that you use. It also used to apply to audio recording too. Keeping the signal away from the noise floor. Although I think the philosophy may have changed with 24-bit recording.

I remember a post a while back about lowering ISO for low contrast situations, and using a higher ISO for high contrast situations. This is because if you use a lower ISO, you will generally open up the iris so it looks right. All you're essentially doing though, is exposing to the right when you're doing this. It's the same as exposing to the right, and bringing it down in post later.

Shooting underwater doesn't change any of this because of low light situations. Adding a low dynamic range feature is not going to increase the sensitivity of the sensor.

Justin Anderson
12-17-2008, 12:31 PM
Well, say one is shooting in low contrast conditions and has a 8-bit AD converter in the camera, but the editing system has higher bit depth. In this case I guess it makes sense to shoot to the right, that is, set the exposure just under clipping (or a certain point of the gamma curve), and then bring it all back to the correct level in edit. As a result one may end up having more data/bits on the mid-gray level. (The rationale of this depends very much on the gamma-curve employed.)

Hm.
I was just working on a shoot on a cloudy day. In retrospect, maybe we should have used a gamma that crushed the lows and the highs.

Lauri Kettunen
12-17-2008, 01:23 PM
In retrospect, maybe we should have used a gamma that crushed the lows and the highs.

Sounds like the most sensible thing to do in such conditions.

Joshua Brown
12-17-2008, 11:45 PM
Overcast (when not using other forms of light control) in my experience has been a very good time to shoot during the day. But like you said, you should balance the histogram toward the middle. But of course each situation differs. Your best bet is to properly balance your monitor (when not using actual light meters) and judge both it and the histogram.

-Josh