In camera "raw histogram"?

[davidedric]

Im my cameras, and so far as I know in others, the histogram displayed in camera references the derived jpeg. This means that there is normally some headroom if you are shooting raw, by which I mean you can increase exposure without actually blowing the highlights when you come to work on the raw file in whatever software you use.

Could there be such a thing as a "raw histogram" displayed in camera to take away the estimating of how much headroom you have?

[xpatUSA]

In my cameras and, so far as I know, in others, the histogram displays in-camera referred to the derived jpeg. This means that there is normally some headroom if you are shooting raw -by which I mean you can increase exposure without actually blowing the highlights when you come to work on the raw file in whatever software you use.

Could there be such a thing as a "raw histogram" displayed in camera to take away the estimating of how much headroom you have?

Hello Dave.

Some years ago an article by Guillermo Luijk addressed that issue. See:

http://www.guillermoluijk.com/tutori...b/index_en.htm

A lot of work developing the necessary Custom WB and it didn't work too well on my Sigma SD10 camera. Haven't tried it since as even raw Foveon data with the much-vaunted "raw headroom" can still blow a channel during conversion to RGB.

P.S. the link he gives for "Show Image" is bad. I do have it on my HD or maybe one of images linked to at the end would work for you ...

[DanK]

You can come reasonably close with Canon bodies by using the neutral or faithful picture style. If I understand, they only differ in terms of WB.

https://digital-photography-school.c...icture-styles/

[Polar01]

A good read that I found was The Optimum Digital Exposure, it explained how to get the optimum exposure for your camera. I found that I could usually shoot about 1-1/3 to 2/3 stops pass what the blinkies were showing and not lose detail in the whites.

Cheers: Allan

[dje]

Hi Dave

Many people have asked for an in camera raw histogram but I very much doubt that camera manufacturers will come to the party. The raw histogram alone doesn’t give you the full story as far as clipping is concerned. For a raw file to be any use, it needs to be processed into a viewable image and a significant part of this process is the application of white balance multipliers. As indicated in the article Ted referenced, these multipliers can be up to 2.5 for the red or blue channels. This needs to be accounted for when looking for clipping margin in every channel. I dare say a raw histogram conditioned with wb multipliers could be produced for a particular wb setting but really I think the whole process is likely to get a bit complicated for the average user.

I reckon I can live with the exposure fudge factor approach.

Dave

[revi]

(...)For a raw file to be any use, it needs to be processed into a viewable image and a significant part of this process is the application of white balance multipliers. As indicated in the article Ted referenced, these multipliers can be up to 2.5 for the red or blue channels. This needs to be accounted for when looking for clipping margin in every channel.(...)

But that's mainy an issue when using integers for image treatment. Some programs using floating point do not have that issue.

In those cases, the raw file data are mapped to a 0..1 float range. At the end of the editing, the float range is mapped back to the requested integer range. Crucially, in between those steps, you won't cause highlight clipping with normal operations. Specifically, exposure correction and white balancing won't cause clipping.

So a pure raw histogram (with no treatment of the data at all) can be useful.

As for the headroom available, that seems to depend on the camera brand, or rather the conversion curve they apply to go from the linear raw data to the log-transformed jpeg data. Those curves also do a form of tone mapping, and some brands are much more aggressive with the highlights than others. And the more aggressively they add contrast, the more headroom you have for raw exposure (more or less)...

[xpatUSA]

Originally Posted by dje (...)For a raw file to be any use, it needs to be processed into a viewable image and a significant part of this process is the application of white balance multipliers. As indicated in the article Ted referenced, these multipliers can be up to 2.5 for the red or blue channels. This needs to be accounted for when looking for clipping margin in every channel.(...)

But that's mainly an issue when using integers for image treatment. Some programs using floating point do not have that issue.

In those cases, the raw file data are mapped to a 0..1 float range. At the end of the editing, the float range is mapped back to the requested integer range. Crucially, in between those steps, you won't cause highlight clipping with normal operations. Specifically, exposure correction and white balancing won't cause clipping.

So a pure raw histogram (with no treatment of the data at all) can be useful.

And not just white balance multipliers, Dave. in my Sigma SD9 with it's Foveon sensor, the raw data is converted first to the XYZ color space by a 3x3 matrix. While matrices are used for other sensor types, the Foveon needs large off-diagonal coefficients, particularly for the blue raw channel where it can be 4 times or more going from blue to to Z, irrespective of the following less severe WB matrices.



Z = .887R+(-3.91G)+4.112B

Although it can seen that the green raw channel takes some away from Z, a scene with little green in can be problematic - for example an urban sunset. Although the converter uses floating point, I'm not sure how it helps that situation.

Pardon the mention of an odd-ball sensor ...

If not sure and knowing that some histograms with apparently good raw headroom can still result in a clipped RGB output image, I use RawDigger's 3 or 4 channel raw histogram, albeit after the event.

[pschlute]

So a pure raw histogram (with no treatment of the data at all) can be useful.

But to what degree ?

I take pictures in raw then convert to TIFF (AdobeRGB), then jpeg (sRGB). The histogram changes with each conversion to a different colour space.

Surely it makes more sense if your output is going to be sRGB for web viewing, that you use a histogram in-camera that depicts that, even when shooting raw ?

[xpatUSA]

But to what degree ?

I take pictures in raw then convert to TIFF (AdobeRGB), then jpeg (sRGB). The histogram changes with each conversion to a different colour space.

Interesting - which type of histogram?

Surely it makes more sense if your output is going to be sRGB for web viewing, that you use a histogram in-camera that depicts that, even when shooting raw ?

That's what I do too because I only view and post in sRGB.

[DanK]

The issue is luminance, not color space as such. The choice of a color space isn't sufficient to deal with the problem. The problem is that because the histogram is based on a JPEG thumbnail, it will typically understate the headroom left in the raw file. That's why an approximation is to choose a picture style that does not increase contrast, saturation, etc.

[xpatUSA]

The issue is luminance, not color space as such. The choice of a color space isn't sufficient to deal with the problem. The problem is that because the histogram is based on a JPEG thumbnail, it will typically understate the headroom left in the raw file. That's why an approximation is to choose a picture style that does not increase contrast, saturation, etc.

I think that the good old "headroom left in the raw file" is a myth, and would love to see credible proof that it is not. DPR makes much of this myth by messing with stuff in ACR (as that were the defining app on the planet) and it gets seriously boring after a while ...

[revi]

I think that the good old "headroom left in the raw file" is a myth, and would love to see credible proof that it is not. DPR makes much of this myth by messing with stuff in ACR (as that were the defining app on the planet) and it gets seriously boring after a while ...

If I push exposition a bit, I often get ooc jpegs with clipped skies etc. The corresponding raw is NOT clipped. Seems proof to me that there is some headroom left with my camera (Sony).
It may be different for different brands, of course, as the tone mapping they do is what gives the brand "look" to the ooc jpegs (and Sony seems to like rather high contrast in the midtones, sacrifising highlights; and shadows).

Other indications: the raw converter I use (darktable) provides a series of tone curves ("base curves") to match the in-camera conversion to jpeg. Those curves are labelled with the corresponding camera brand. And some are virtually horizontal in the highlight section, throwing away detail in that are. Others (e.g. "Leica") are less contrasted, and give more room to the highlights.
Note that those base curves are generated from raw <-> ooc jpeg pairs. There's a description of the base curves and how to generate them on the darktable site

[Manfred M]

I think that the good old "headroom left in the raw file" is a myth, and would love to see credible proof that it is not. DPR makes much of this myth by messing with stuff in ACR (as that were the defining app on the planet) and it gets seriously boring after a while ...

Ted - we have been through this before and if I remember correctly, your testing agreed with mine.

I'm with Revi on this one, on my Nikon D810, I get around 1-1/2 additional stops in my raw data. That being said, this is for the base ISO, as I try not to shoot at higher ISO unless I have to. In studio settings, I can pump in enough light for this to wor;. In my landscape work I use a tripod, so a slow shutter speed allows me to use low ISO for most image. I would expect a drop in the raw data headroom if shooting oat very high ISO settings because of a drop in dynamic range. I will have to test this some day, but unfortunately my camera is in the shop right now.

[xpatUSA]

Ted - we have been through this before and if I remember correctly, your testing agreed with mine.

I'm with Revi on this one, on my Nikon D810, I get around 1-1/2 additional stops in my raw data. That being said, this is for the base ISO, as I try not to shoot at higher ISO unless I have to. In studio settings, I can pump in enough light for this to work. In my landscape work I use a tripod, so a slow shutter speed allows me to use low ISO for most image.

Probably so for CaNikOny cameras, Manfred, so I won't argue about that.

I would expect a drop in the raw data headroom if shooting at very high ISO settings because of a drop in dynamic range. I will have to test this some day, but unfortunately my camera is in the shop right now.

On my favorite cameras, the higher the ISO, the more the raw headroom because they are ISO-less. So, 1600 ISO gets me 4EV raw "headroom" but, as you say, less DR and plenty more noise ...

[Stagecoach]

Everything I have read about this seems to come to the conclusion that there are just so many 'variables' it would not give a conclusive indication for every eventuality.

I wonder if some time in the future we may see a 'traffic light' system in the viewfinder for RGB and brightness levels where its gain can be set by the user?

I've just undertaken a 'hack' to my Z6 so that the viewfinder turns any bright area black at a certain level. This at times gives an easier to see indication than the small histogram.