Reading mode: Light Dark

Basic question on exposure via histograms

Go to last post
N

nathanbush

Member
Messages
31
Reaction score
0
Location
Brooklyn, US
I've been lately into darker images (or at least true to the original, if the original was dark). I'm wondering what the best practice is for exposing for the highest quality end product.

Let's take a hypothetical example. Say I underexpose in camera, so the information is all in the darker end of the histogram, but no clipping occurs. Alternatively, I could overexpose, where all the info is in the lighter end of the histogram, but still no clipping. If I'm going to correct the exposure in post (eg, make the darker picture lighter, or the lighter picture darker, to achieve the same exposure), are the quality of the pixels going to be better in one scenario or the other? Are all pixels equal, and equally usable, if they lie within the boundaries of the histogram?
 
1. Noise

The noise is unavoidably higher in the dark regions than with higher exposure. Increasing the intensity in the raw processing ("pushing the exposure") increases the noise as well (the noise remains the same relatively to the intensity), but reducing the intensity in the raw processing does not create noise.

2. The number of different pixel levels, as Peter wrote it in the above post, may be a consideration, though much less with modern cameras, which create many more pixel levels than earlier ones.

Keep in eye, that if you can not increase the exposure, for example the shutter time has to be short because of movement, then increasing the ISO is a good way as well, while keeping the exposure unchanged .

See http://www.cryptobola.com/PhotoBola/SourceOfNoise.htm

--
Gabor

http://www.panopeeper.com/panorama/pano.htm
 
If you're shooting Raw, over-exposing and then adjusting in the Raw processor is pretty much the way to go.

If you're shooting JPEG, be aware that a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness.
 
Doug Pardee said:
If you're shooting JPEG, be aware that a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness.
Click to expand...
Wouldn't this be a good application for opening the JPG in ACR and using the "Exposure" slider to reduce brightness? Wouldn't that undo the gamma change?

Yeah, it would be best to use RAW, but if you did use this technique with JPEGs, isn't ACR's Exposure slider the most appropriate tool to re-adjust exposure?

Wayne
 
Wayne Larmon said:
Doug Pardee said:
If you're shooting JPEG, be aware that a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness.
Click to expand...
Wouldn't this be a good application for opening the JPG in ACR and using the "Exposure" slider to reduce brightness? Wouldn't that undo the gamma change?
Click to expand...
1. Doug must have been thinking of the "S curve", not the gamma curve. The gamma curve is a quite straightforward issue (it depends on the color space). However, the S-curve is an optional adjustment, depending on the contrast setting. There is no way to know, if and how exactly that was applied.

2. Even if no S curve has been applied, following problem exists: the number of RGB levels in the very highlights is low (there is not much contrast there). If the "brightness" has to be reduced, then the same range of the original intensities requires much more RGB levels than there were stored in the highlights. The consequence is, that the area with the reduced intensity will have very low contrast.

If an S curve has been applied (which is the normal case), then this becomes much worse, because that caused even lower contrast at the very bright end, reducing the number of stored levels.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm
 
If your camera supports it, be sure to use RGB histograms instead of the single luminosity histogram. I had some bad shots of very saturated red flowers where the luminosity histogram gave no indication that I was clipping in just the red channel. The result was banding in the bright red areas of the image.
--
A pixel is a terrible thing to waste.
 
The "clipping" you have seen was caused most probably by the color space. When processing the raw, specify ProPhoto RGB and the clipping will vanish.

Of course that does not help you, for you need the final image in Adobe RGB or sRGB. However, if the clipping did not occur in the raw channels - as I suspect it - then you need to reduce the brightness to remain in sRGB, but the image remains cleaner in the shadows, better than if you reduce the exposure when shooting.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm
 
GaborSch said:
1. Doug must have been thinking of the "S curve", not the gamma curve. The gamma curve is a quite straightforward issue (it depends on the color space).
Click to expand...
I was thinking of the gamma curve. It is a straightforward issue only if you're knowledgeable about such things and have the proper tools, which most people aren't and don't.

Most photo editing software treats R'G'B' as being RGB, and scales it directly without regard to the gamma curve. If you know that your software properly deals with the gamma curve, then everything's fine. Personally, I use LightZone, which makes a big deal about doing its adjustments in a linear color space. I imagine that there are other tools that do so. Using a linear color space in Photoshop would work, too, as long as you know to do so.
 
The conventional wisdom seems to be "expose to the right," for the reasons spelled out above. How far to the right is the question. For most purposes clipped highlights detract more from the photo than shadow noise, and with many camera histograms it can be difficult to see just where clipping sets in.

I suggest using the 4-color histogram and being a bit conservative unless you really know the camera well. With many cameras it is very possible to clip one color without it showing on the luminance histogram. Blinking highlights can also be useful.

Gato

--
Street Fashion and Alternative Portraits:
http://www.silvermirage.com
 
Carey Brown said:
I stand by what I said. Even Gato's comment backs me up.
Click to expand...
"Standing by" is hardly an argument. I base my statement on the analysis of thousands of raw files from five dozen camera models, on raw level (what you probably never see).

Anyway, it's your business, but it's a pity to ruin good shots by underexposing them due to lack of understanding the issues.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm
 
It's not under exposing if you expose to the right. It's just that you can't trust the luminescence histogram to tell you the whole story. I'm sure that with all your experience you've looked at the histograms in both luminescence and RGB mode and you've seen the difference.
--
A pixel is a terrible thing to waste.
 
Carey Brown said:
If your camera supports it, be sure to use RGB histograms instead of the single luminosity histogram. I had some bad shots of very saturated red flowers where the luminosity histogram gave no indication that I was clipping in just the red channel. The result was banding in the bright red areas of the image.
Click to expand...
One additional reason why camera histograms can be inaccurate is because most cameras base the histogram on JPEG data. Even if you are shooting RAW.

Some have found that setting your camera's saturation and contrast to the minimum makes the histogram be more accurate (if you are shooting RAW). See

Settings for an Accurate Histogram
http://www.luminous-landscape.com/tutorials/right-hista.shtml

I've found that when I do this, that the camera histogram does better match the histogram that I see when I load the images into ACR.

Carey, you can test Gabor's statement about color spaces by going back to images that show clipping in the red channel and re-convert them using the ProPhoto RGB color space. This isn't a theoretical issue: either your RAW files clip in ProPhoto RGB. Or they don't.

And I agree with you that the RGB histograms are more accurate than the single luminance histogram. If the luminance histogram showed the maximum of R or G or G, then it would be accurate, insofar as showing you when any channel is clipping. But it doesn't.

Wayne
 
Carey Brown said:
It's not under exposing if you expose to the right. It's just that you can't trust the luminescence histogram to tell you the whole story. I'm sure that with all your experience you've looked at the histograms in both luminescence and RGB mode and you've seen the difference.
Click to expand...
The color histogram is not more reliable than the luminance histogram. They represent the JPEG image (which is embedded in the raw file). True ETTR can be achieved only with a special, neutral setup, including a neutral custom WB; see

http://forums.dpreview.com/forums/read.asp?forum=1037&message=26905476
http://forums.dpreview.com/forums/read.asp?forum=1019&message=26534648
http://forums.dpreview.com/forums/read.asp?forum=1019&message=27554522

There are threads regarding the imaginary clipping; typically, Nikon's newer cameras, the D300, D3/D700 have often shown this problem. One of those threads with detailed discussion is http://forums.dpreview.com/forums/read.asp?forum=1039&message=28668040

These posts contain the raw histograms of images, which caused red clipping in the raw conversion :

http://forums.dpreview.com/forums/read.asp?forum=1039&message=28702955
http://forums.dpreview.com/forums/read.asp?forum=1039&message=28727959

--
Gabor

http://www.panopeeper.com/panorama/pano.htm
 
Doug Pardee said:
GaborSch said:
1. Doug must have been thinking of the "S curve", not the gamma curve. The gamma curve is a quite straightforward issue (it depends on the color space).
Click to expand...
I was thinking of the gamma curve. It is a straightforward issue only if you're knowledgeable about such things and have the proper tools, which most people aren't and don't.
Click to expand...
Doug, earlier you said:

" If you're shooting JPEG, be aware that a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness. "

Are you familiar with Adobe Camera RAW (or its Lightroom incarnation)? If so, then am I correct that opening a JPEG image in ACR and using the Exposure slider to reduce brightness will account for the gamma issue?

If you aren't samiliar with ACR then you should know that Adobe has pulled a rather large rabbit out of the hat. ACR is nominally a RAW converter. But the most recent versions of ACR let you process JPEG files using the same controls that are normally used to control RAW conversion. Note that this implies a lot of emulation and tricky transforms. Comparable in complexity to soft proofing. Simple on top, but lots going on under the hood.

If my understanding is correct, then, if processing a JPEG in ACR, then the Exposure slider is applying the same gamma curve to the JPEG that it does to a RAW file. This is the question I originally asked you.

But answering it requires understanding how ACR processes JPEG files. I'm guessing how it works. I'd like to know if my guess is accurate.

Wayne
 
Comments as follows:
Wayne Larmon said:
Doug Pardee said:
GaborSch said:
1. Doug must have been thinking of the "S curve", not the gamma curve. The gamma curve is a quite straightforward issue (it depends on the color space).
Click to expand...
I was thinking of the gamma curve. It is a straightforward issue only if you're knowledgeable about such things and have the proper tools, which most people aren't and don't.
Click to expand...
Doug, earlier you said:

" If you're shooting JPEG, be aware that a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness. "

Are you familiar with Adobe Camera RAW (or its Lightroom incarnation)? If so, then am I correct that opening a JPEG image in ACR and using the Exposure slider to reduce brightness will account for the gamma issue?

If you aren't samiliar with ACR then you should know that Adobe has pulled a rather large rabbit out of the hat. ACR is nominally a RAW converter. But the most recent versions of ACR let you process JPEG files using the same controls that are normally used to control RAW conversion. Note that this implies a lot of emulation and tricky transforms. Comparable in complexity to soft proofing. Simple on top, but lots going on under the hood.

If my understanding is correct, then, if processing a JPEG in ACR, then the Exposure slider is applying the same gamma curve to the JPEG that it does to a RAW file. This is the question I originally asked you.

But answering it requires understanding how ACR processes JPEG files. I'm guessing how it works. I'd like to know if my guess is accurate.
Click to expand...
Wayne, it seems to me that ACR can't know exactly what Tone Response Curve (TRC) was applied to the JPEG and even if it reverses out a response curve, applies the correction, and then reapplies the TRC unless the image was originally created with ACR , ACR can't know what the TRC was from metadata and would just have to use a generic one such as gamma 2.2 or the sRGB TRC.

Further, it would seem that one could test for this by creating a JPEG with a rather extreme TRC, both generated outside ACR and with ACR, to see what happens when exposure adjustments are made.

Regards, GordonBGood
 
Doug Pardee said:
If you're shooting Raw, over-exposing and then adjusting in the Raw processor is pretty much the way to go.

If you're shooting JPEG, be aware that a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness.
Click to expand...
I don't understand this. Can someone explain? Thanks.

"...a gamma curve has been applied to the data values, and most photo editing software doesn't take that into account when adjusting lightness."

Never mind. I see this was addressed in other comments.
 
GaborSch said:
Carey Brown said:
It's not under exposing if you expose to the right. It's just that you can't trust the luminescence histogram to tell you the whole story. I'm sure that with all your experience you've looked at the histograms in both luminescence and RGB mode and you've seen the difference.
Click to expand...
The color histogram is not more reliable than the luminance histogram. They represent the JPEG image (which is embedded in the raw file). True ETTR can be achieved only with a special, neutral setup, including a neutral custom WB; see

http://forums.dpreview.com/forums/read.asp?forum=1037&message=26905476
http://forums.dpreview.com/forums/read.asp?forum=1019&message=26534648
http://forums.dpreview.com/forums/read.asp?forum=1019&message=27554522

There are threads regarding the imaginary clipping; typically, Nikon's newer cameras, the D300, D3/D700 have often shown this problem. One of those threads with detailed discussion is http://forums.dpreview.com/forums/read.asp?forum=1039&message=28668040

These posts contain the raw histograms of images, which caused red clipping in the raw conversion :

http://forums.dpreview.com/forums/read.asp?forum=1039&message=28702955
http://forums.dpreview.com/forums/read.asp?forum=1039&message=28727959
Click to expand...
Great info ono these threads. Thanks. Mary
 
You must log in or register to reply here.

Similar threads

B
Default exposure compensation method is different on Sony - pluses and minuses
Replies
4
Views
8K
A-Marc
A-Marc
M
What new computational features would you like to see?
2 3 4
Replies
61
Views
18K
Muster Mark
M
lokatz
Noise Reduction: Comparing DxO and LRC
2
Replies
31
Views
28K
Vince P
Vince P
B
Question about Sony RAW through Lightroom Mobile to HDR.
Replies
2
Views
15K
dalethephotographer
D
OutsideTheMatrix
FZ300 zebra stripes
Replies
6
Views
498
OutsideTheMatrix
OutsideTheMatrix

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top