Exposure and image quality

[bernard_yomtov]

Suppose I shoot at 200 ISO, say, and underexpose by two stops, in RAW format. Then,

when I download the file I put the two stops back in by applying exposure compensation.

Will my image be better, worse, or the same as if I just shot at 800 to begin with?

 

I'm using a Nikon D70, if that makes a difference.

[dean_g]

Seems like a good question. I've done that with cameras that don't have higher ISOs but I've never compared on a camera that offers higher sensitivities. The results will be probably be different, but better or worse.. I assume you mean more or less visible noise.. but there's probably other things like effects on color or color saturation.

 

I'm going to try it.. you should too. It's free!

[._kaa]

The image shot at ISO 800 will be better, considerably, than an image shot at ISO 200 and pushed two stops at the RAW conversion stage.

 

That is assuming a "normal" image with important shadows, since it's the shadows that will suffer the most during pushing.

[michael_taylor9]

I would assume that it's similar to shooting film in that you can't "put detail back" that wasn't recorded in the first place. With film if I push the film by shooting 2 stops higher than the rated ISO I would have the lab pull the processing time to reduce the added contrast, thus preserving the details in the shadows and the highlights.

I'm still not convinced that going digital is the way to go for me yet because of things like this problem being discussed here. At least with film I know where I stand and what I can expect because I am controling the entire process from capture to print. There's my 2 cents.

[jason_reed2]

This question certainly sparked some interest, so I did a quick and dirty.

 

http://www.metroaffinity.com/isoexposuretest/

 

The results were closer than I expected, but it's still pretty clear the quality at ISO 800 is better.

[._kaa]

<p><i>At least with film I know where I stand and what I can expect because I am controling the entire process from capture to print.</i>

 

<p>LOL. To control "the entire process" with film you need to develop the film at home and print it at home, too. Very few people do this.

 

<p>With digital, on the other hand, I am actually in full control of the entire process -- from RAW conversion to white balance to contrast/saturation/etc. tweaks. Even if I send off my images to a lab to be printed, I always specify "no adjustments, print as is" and it works pretty well :-)

 

<p>So I would argue that for the great majority of people digital provides more control over the process than film.

[godfrey]

<i>.."Suppose I shoot at 200 ISO, say, and underexpose by two stops, in RAW format.

Then, when I download the file I put the two stops back in by applying exposure

compensation. Will my image be better, worse, or the same as if I just shot at 800 to begin

with?" ...</i>

<br><br>

I've tested this thought exercise with several cameras.

<br><br>

You would be better to set ISO 800 and expose properly (which in digital sensor terms

means exposing as generously as you can without hitting the hard clip at sensor

saturation in highlight areas), saving the exposure in RAW format, and then processing the

image to bring up the shadow areas to the greatest extent possible.

<br><br>

Underexposure on a digital sensor means losing shadow detail to noise, and no amount of

"push" post-processing will bring it back. You can retrieve much more detail from slightly

overexposed (but not complately saturated) highlights than you can from overly

underexposed shadows.

<br><br>

Godfrey

[beeman458]

"Will my image be better, worse, or the same as if I just shot at 800 to begin with?"

 

Dumb question. What were your results when you tried this yourself?<p>

 

My expectations, you're gonna force noise into the shadows by doing this because you'll unnecessarily load up your shadows and in doing so, will take flexibility away from your highlights by compressing them.<p>

 

Do you <a href="http://www.sportsshooter.com/special_feature/chimping/index.html">chimp</a> your <a href="http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml">histogram?</a><p>

[roger krueger]

<i>Underexposure on a digital sensor means losing shadow detail to noise, and no

amount of "push" post-processing will bring it back. You can retrieve much more detail

from slightly overexposed (but not complately saturated) highlights than you can from

overly underexposed shadows.

</i><br><br>

On what camera? This exactly the opposite of what I see on my 1ds II--on raws I can dig

deep into the shadows, but highlights are very easy to blow out past all repair.

 

The advantage I see shooting underexposed at lower ISOs is that the shadow noise is only

slightly worse, the highlight noise is somewhat better, and you've extended the useful

dynamic range--normally my Canon has about 4 stops of shadow lattitude, but maybe a

half stop of highlight lattitude. Shifting that to be more balanced is useful in high-scene-

contrast situations.

[petar_lambrev]

Bernard, thank you for asking this question. I've been wondering the same thing for some time but I haven't got the answer yet.

 

I believe that it is virtually the same whether you choose a higher ISO setting or underexpose and correct the RAW upon conversion. I think the ISO setting is actually electronic Gain applyied to the sensor and electronic gain does not improve the signal to noise ratio - it merely amplifies the signal along with the noise.

 

So, regarding the signal-to-noise ratio it shouldn't really matter if you apply this gain in-camera or post-shooting. There might be some advantage in setting the proper ISO because the gain is applied before sampling (digitizing) the intensity of each pixel. But I think the sampling resolution of DSLR sensors (12 bits or more) is sufficiently high so that you won't notice the difference as the examples above show.

[beeman458]

"But I think the sampling resolution of DSLR sensors (12 bits or more) is sufficiently high so that you won't notice the difference as the examples above show."

 

Wiiiiiine! (with nasal ephesis)

 

If you above is so, then why do my images (shadows) get destroyed when I turn up the exposure in post or in conversion?

 

My suggestion, give it a try so you folks can see what happens to the quality of your shadows.

[nicholas_king1]

The 800 iso will a image with more depth as the right side of the histogram contains a greater data density this is due to non-linear nature of photo sensors

 

 

A 12 Bit raw File

 

Within the first F/Stop, which contains the Brightest Tones 2048 levels available

Within the second F/Stop, which contains Bright Tones 1024 levels available

Within the third F/Stop, which contains the Mid-Tones 512 levels available

Within the fourth F/Stop, which contains Dark Tones 256 levels available

Within the fifth F/Stop, which contains the Darkest Tones 128 levels available

 

More details:

 

http://www.luminous-landscape.com/tutorials/understanding-series/u-raw-files.shtml

[petar_lambrev]

Thomas, nobody insists that a camera can record infinitely deep shadows. At some point inevitably the noise will get higher than the signal and the shadows will simply not be there.

 

I was just speculating that it is quite possible that the signal-to-noise ratio is not getting any better with increasing the ISO so you won't capture more of the shadows really. I may be dead wrong about that and I'm looking for the correct answer. And it might also depend on the camera model as well.

[beeman458]

"Thomas, nobody insists that a camera can record infinitely deep shadows. At some point inevitably the noise will get higher than the signal and the shadows will simply not be there."

 

I'm not saying that anybody is insisting anything other then their reluctency to set a camera on a tripod and take eight equivalently exposed images. One image at a proper exposure of ISO 100, ISO 200, ISO 400, ISO 800 and the same ISO settings, with exposure compensation set at a minus two stops and see for themselves what happens. Seems like an easy enough request.

 

Myself, I'm just reporting that based on my experience, when I try to do more than about a half stop correction, the shadows are destroyed.

 

"Will my image be better, worse, or the same as if I just shot at 800 to begin with?"

 

In regard to the original question, the answer, the image will be worse. I was trying to encourage everybody to try for it themselves, cause the exercise/information gained is a real time benefit as opposed to asking someone and simply reading their response.

 

I posted a link to an article which gave information about the why's of the histogram and the non-linear nature of how exposure information is distributed. In the case of a properly exposed image, you can, without destroying the image, expand the information. But if you underexpose the information by two stops, then the net result will be clipping of the shadow information and once clipped, it's gone. Period!

 

Hope my above lends clarity to the issue.

[beeman458]

As to the cameras and the sensor question...... The 20D, 1DMkII and the 1DsMkII are better at dynamic range issues then the previous generation, the 10D, 1D and the 1Ds by about a stop or so.

 

But still, once the information is clipped, it's gone.

[beeman458]

When you jam the sliders, watch, you'll destroy the contrast and when you try to call the contrast back, D'oh!, there goes the information, right before your eyes.

 

No free lunch.

[petar_lambrev]

Thanks, Thomas. I followed your links in the first place, they were not new to me. I understood your suggestion for everyone to try for themselves. I don't possess a DSLR and couldn't do that. I don't question your experience and I take your word that underexposing is worse without doubt. I'm just trying to explain it for myself. I still believe that you don't improve the SNR with higher ISO and don't capture more shadows (nobody answered this particular question). If this is correct, then the underexposed images must be ruined because of the reduced dynamic range (which is because the 2 stops gain was applied after sampling and not before as with setting the ISO). I know how pumped 8-bit JPEGs look like. I thought 12-14 bits in the RAW would be enough to not notice the loss. Apparently wrong.

[godfrey]

<i>GDG: .."You can retrieve much more detail from slightly overexposed (but not

completely saturated) highlights than you can from overly underexposed shadows."...

<br>

RK: "On what camera? This exactly the opposite of what I see on my 1ds II--on raws I can

dig deep into the shadows, but highlights are very easy to blow out past all repair."... </i>

<br><br>

The cameras I've worked with that have RAW format storage capability are the Konica

Minolta A2, Canon 10D, Olympus C8080WZ and Pentax *ist DS.

<br><br>

The problem of highlight blowout was not ignored ... as I said, "slightly overexposed but

<b>not completely saturated</b> highlights" can be worked with easily. Slightly

overexposed means hitting highlight sensor values that you want *some* detail in around

the 3700 mark, in 12 bit space (210-230 in 8bit space). Underexposed

shadow values, well, it depends upon how underexposed you go, as noise intrusion is a

"soft" limit compared to the "hard" saturation clipping limits on the highlights. That's a

critical boundary.

<br><br>

Remember that in a 12bit per pixel RAW image, the sensor data is a linear gamma with

4096 levels. Human vision is emulated with a gamma curve that compresses highlight

values together and expands values in the shadow areas from that linear

gamma quantization space. More than half the sensor's quantization space is devoted to

highlight values that are compressed into a quarter to a third of the RGB channel output

after gamma conversion. This happens in the RAW conversion process. If you use an

exposure which just avoids the saturation clipping of highlights, you change the

gamma curve and retrieve much better shadow values, with greater separation and more

data, without sacrificing highlight detail.

<br><br>

For people still thinking in terms of film characteristics: <br>The sensor's linear gamma

capture is completely different from film capture, because film

emulsions have been designed to emulate human vision's compression/expansion gamma

curve.

<br><br>

Godfrey

[beeman458]

"Suppose I shoot at 200 ISO, say, and underexpose by two stops..."

 

The question had to do with underexposure by two stops, not slightly overexposed.

 

We can go into slightly overexposed and show what that does to an image as well if you're so inclined. Not trying to be argumentative but digital is very, very persnickety; unforgiving too:)

[godfrey]

You're entitled to your opinion, Thomas. Someday, though, you might understand

something and change it.

 

Digital capture is different from film capture, if you think in film terms you will not

understand "proper exposure" with a digital sensor. That's all. The response I made

addressed what my experiments have demonstrated regards the quality achievable by

underexposing by two stops and "pushing" the result vs using a higher sensitivity and

exposing correctly.

 

Godfrey

[beeman458]

"You're entitled to your opinion, Thomas. Someday, though, you might understand something and change it."

 

Sorry, it's not opinion but a fact that's replicatable, hence why I suggested trying to do what I suggested so one could see for themselves and not have to take my word for it.

 

"Someday, though, you might understand something and change it."

 

After almost forty years with film, LF, MF and 35mm, Zone System training, digital and a pro-degree, I have a clue but you too are welcome to your opinion:)

[beeman458]

"if you think in film terms you will not understand "proper exposure" with a digital sensor. That's all."

 

Just curious, where am I referring to film in my postings or the links I left?

[godfrey]

Thomas, <br><br>

All of your posts indicate a strong theme of film background and expectations. There's

nothing wrong with that. But your notions about the sensitivity of digital image exposure

come out of that. In truth, it is neither more nor less "finicky". They're simply

different, as it is a different medium.

<br><br>

(Color balance on a wet lab print process, now that's freekin' finicky. :-)

<br><br>

I also have been doing photography for 40+ years with similar kinds of equipment and

training experience; I feel I can sense the thought processes behind your statements

because they are similar to my own.

<br><br>

best,<br>

Godfrey

[beeman458]

"All of your posts indicate a strong theme of film background and expectations."

 

If it helps, all my comments today are based upon my digital experience and none of my film experience.

 

"But your notions about the sensitivity of digital image exposure come out of that."

 

In film, with my eye in the grain magnifier, it killed me to watch the image tear up. In digital, on the monitor, same thing. It just kills me to watch detail get destroyed with contrast, level and saturation settings. Love Genuine Fractals and love tweeking an image with selective masking:)

 

"In truth, it is neither more nor less "finicky". They're simply different, as it is a different medium."

 

Can't argue that point it's just that I'm currently all digital so my focus is 100% digital in my thinking.

 

"(Color balance on a wet lab print process, now that's freekin' finicky. :-)"

 

Don't even want to talk about filter packs:) Thank God for a densitometer cause I was dead without one to read the inserted gray card corner:)

 

Thanks for your thoughs.

 

Just a word to the wise, stay away from the flue thats going about, comes in like a freight train and hangs all the next day with a 102F temperature. I can see why folks are ending up in the hospital with this stuff.

 

"I also have been doing photography for 40+ years with similar kinds of equipment and training experience; I feel I can sense the thought processes behind your statements because they are similar to my own."

 

Oops! Busted:) Like your picture of the week idea and the content.

[godfrey]

<i>.."Just a word to the wise, stay away from the flue thats going about, comes in like a

freight train and hangs all the next day with a 102F temperature. I can see why folks are

ending up in the hospital with this stuff.".. </i>

<br><br>

uh huh. Had that the last week of January. Took me out for 5 days total. Bleck.

<br><br>

Godfrey