Can you tell the difference between an image shot natively at ISO 12800 and one underexposed at ISO 6400 and pushed +1EV in post?
Updated: Now includes samples converted with the RAW converter RPP.
The common assertion is that an image with software-applied “gain” should be inferior to the image with in-camera processing with regard to noise. And logically, that makes sense. How can an “underexposed” file possibly compete for quality with an image that has been “properly” exposed?
In using the Nikon D3 for concert photography, tweaking the exposure compensation of an image in post is a quick and common adjustment in my work flow. While most of the time adjustments if any are small, larger changes may come into play, whether as outright damage control or by design.
But going back to what logic tells us, isn’t this digital “push processing” wrecking havoc on noise levels? I performed a quick test of the D3 to provide some samples for this debate.
Here’s the full-frame of the test images:
The Setup
Both images were shot with the Nikon D3 and Nikon 24-70mm f/2.8. The scene was illuminated by an SB-900 mounted on-camera and bounced off the ceiling.
The shutter speed, aperture, and flash output of both shots remained the same; only the ISO was changed between frames, resulting in one-stop underexposure for the ISO 6400 shot at the time of capture.
Both images were processed from RAW in Nikon NX2 with in-camera, high ISO noise reduction disabled. Standard sharpening was applied (level 3) to both images as shot in-camera.
The Crops
Below are 100% crops from the two files. The sample A crops in both examples were taken from one image, the sample B crops taken from the other.
Update
Just for a slightly different take, I’ve included two samples from the same RAW files, but this time processed with RAW Photo Processor. All settings for RAW conversion remained the same except +1ev to the ISO 6400 file. I always find that RPP has an interesting take on RAW files, rendering with very smooth gradations and a very tight noise pattern.
There is a slight shift in focus in the above samples, but this does not affect noise levels.
The Pepsi Challenge
So, do you want to play? Which sample is from the native ISO 12800 image, and which is from the digitally-pushed ISO 6400 file?
Do you perceive a consistent difference between the images, and if so, which do you prefer?
The results of this test should be taken to be specific to the Nikon D3 and Nikon Capture NX2 software only. Other combinations of cameras and/or RAW software may render different results based on the implementation of gain at high ISO.
This entry was posted on Monday, January 26th, 2009 at 11:33 pm and is filed under Photography Gear and tagged with analog, d3, digital, gain, high iso, nikon, push processing. You can follow any responses to this entry through the RSS 2.0 feed.
Welcome to ishootshows.com, your source for music photography featuring daily concert coverage, portrait features, and photo advice from professional music photographer Todd Owyoung.
Support this site by purchasing your photo equipment through my affiliate links to B&H Photo Video or Amazon.com
Learn more about how you can support the features of ishootshows.com.
Shooting Caitlin Rose and band before the Old 97s was a fantastic warm up, and a nice way to dust of…
Here’s a little preview from a recent shoot with my friends Option Control toward the end of l…
Today Adobe has released the beta version of Lightroom4, the successor to their popular image manage…
You’ve seen it before – weird, colored blobs in your photographs radiating outward from some…
About that time when I said about the megapixel war being over with the release of the Nikon D4 and …
Why not play i say! I’m going to take my 50-50 coin flip and say “A” is the pushed version, and “B” is the 12800, noting that I haven’t played around wit the D3 to have a full appreciation of how both perform.
Hey Joel,
Nice to hear from you, thanks for playing! Do you prefer one set of samples over the other?
On the small sample size, yep. A seems to provide better depth in detail the image (sorry mind sucks, can’t quite explain it) – the dial + the shadow areas around where you’d grasp the camera are the two bits in particular.
Hiya
I think B looks to be lower quality.. so I’m going to go for that as the pushed version..
Hey Kristen Joel, thanks for your thoughts on this test, I appreciate it.
While there is a little split on exactly which samples are from which files, both of you prefer the quality of the A crops.
I’m with Joel, I think that A is the pushed version. The detail in the writing looks sharper, which says to me that the pixels were smaller to begin with. I’d also prefer A, for the same reason of the sharpness.
Hmmm.
Thanks for your thoughts on these images, Tasha. That’s an interesting note about the perception of detail in A.
The plane of focus did shift minutely, but hopefully this isn’t having a major impact on the way the different samples are read.
I think the most noticable difference is that the B crop has more chroma noise but I pretty much only notice it in the black areas (body of the camera). Other differences are very slight to my eye and might be explained by the focus shift.
Hey Karl, thanks for weighing in here.
The differences are subtle, but I think you’re right that there is some differentiation in the character of the noise in the two samples.
Hey there Todd!
My guess would be that A was shot at native ISO 12800 and B was shot at ISO 6400 and pushed 1 stop in post-processing, but I may be wrong (I probably am :p). The differences between the photos are so subtle that whatever guess I make will be a rather moot point.
Hey João, great to hear from you, thanks for playing! May I ask if you prefer one sample over the other, or otherwise why you guessed as you did?
I’ll say A is the 12,800.
I’m hoping digital processing wreaks havoc.
I can see more depth in the B results, particularly in the grooves on the silver dial and the L on the camera body.
There is a slight amount of more noticeable noise in B but not enough to go “Hey A is so much better I’ll shoot (the A way) way as it creates less noise”.
Hey Matthew, thanks for your thoughts and reasoning of your guess. Good stuff.
Cool test by Todd – High ISO vs. processing: http://tinyurl.com/aatw43
And Todd’s photography rules, so check him out anyway.
I’m currently on my laptop with a bad monitor, but from I can tell I would give A a slight advantage. It seems to have slightly finer texture in the noise.
I have two questions about this test:
1. Wouldn’t the difference be two stops if you used the same shutter speed and aperture and doubled the ISO setting as well as doubling the flash output?
2. The shots both used the native ISO 6400 of the sensor. Both were then pushed, only one of them was pushed by software in the camera while the other was pushed in post-processing. Wouldn’t it be more fair to compare ISO 3200 and ISO 6400 on the D3, as well as using the same lighting and only changing the ISO sensitivity?
If I would have to guess, I would say that A was shot pushed to ISO 12800 in the camera due to the doubling of incoming light that it received if I interpret your post correctly.
Hey Jonas,
Thanks for the feedback. I you have time to check out the test on a different monitor, I’d love to hear if your opinion changes!
1) You’re right! I believe the difference in flash power came from when I was initially looking at a reference shot at ISO 6400 that was “properly” exposed. The flash output for both was 1/128 from the SB-900, good catch. The ISO 6400 shot therefore received one-stop underexposure, all other factors being equal.
2) You’re right in that both shots are “pushed,” though I don’t believe Nikon has ever specified how it achieves gain at high ISO and, in particular, extended high ISO. As far as I know, at the very least, the extensions are stated as such because there is no promise that they are truly calibrated. There are some who say that digital, in-camera gain is applied even earlier on the D3. I may do a full test with a range of ISO at some point.
I should add that a reader’s comment sparked the idea for this test. Their claim that “There is a lot more noise when actually setting the camera to 12800″ when pushing an underexposed ISO 6400 shot via software. Hence the particular settings used for this test. Thanks for your thoughts on this.
Updated: new RAW converter in the mix (re:Can you tell the difference between ISO 12800 and ISO 6400 +1ev?): http://is.gd/hmaV
I’d guess the same as Matt Miller above.
Great test, though!
Thanks, Steve. I’ll tally up the votes and preferences shortly before the answers are given.
Push-processing tally: 4 votes for sample A as pushed, 6 votes for sample B as pushed, 3 votes undecided/no difference http://is.gd/hmaV
I have to say, I don’t see any significant difference between the two samples. I think a lot of the above comments are coming from the difference created by the slight focus shift. At best I would say that sample A shows slightly increased contrast resulting in greater apparent detail. RPP does a remarkable job retaining detail and producing much tighter noise, even if the resultant images look darker, and in this test has proven itself to be the superior raw converter for high ISO images. But for all practical purposes, I say there is no difference between the two samples. If I have to pick, my coin flip says A is native 12800.
Hey Joey,
Thanks for your thoughts on this. I would agree in saying that the focus shift that occurred may be influencing people’s opinion of detail in certain areas of the image. Sample A’s point of focus is slightly more forward than that of sample B. Still, the character of the noise is of course unaffected by focus.
You do bring up an excellent point of the contrast, which has been hinted at before. There is a very minor difference in contrast between the two samples, and you’re right, this does affect perceived detail.
RPP really is an amazing RAW converter – you should try it if you use a mac. On both low ISO and high ISO images, it renders a staggering amount of detail, as well as very nice tonal gradations and fine noise, as noted. Compared to the output from Nikon NX (or just about any other converter), the files it outputs look as if they ripped the AA filter from the camera. But that’s for another post. :)
Thanks for your vote and insightful comments, much appreciated.
I’m saying B is pushed in post.
Thanks for the vote, Woody. I’ll be posting the answer shortly, as well as a brief follow up for the reason for the similarity of these two shots, which is all about two words: unity gain.
I’m also going to say A is pushed, based on my experience (although not knowing what a D3 shot looks like at 100%, I’m guessing to come degree). I also prefer the small version of A because it seems to have a bit better contrast/black level, but that might change at a larger size.
Hey Doug, thanks for your comment. You’re not alone in your thoughts of the difference in the contrast of the two samples.
Cool test by Todd – High ISO vs. processing: http://tinyurl.com/aatw43
And Todd’s photography rules, so check him out anyway.
I can’t say which is which, and I’ve only read the first dozen or so comments in this thread, but I notice a quite a difference in the clarity of the black and white text. It’s consistently crisper and less fuzzy in B.
Hey Paige, thanks for your thoughts!
I can’t guess…its too hard…I dont want to be wrong…ok…
THE LEFT ONE IS PUSHED? Thats my guess!
Hey Keith,
It’s supposed to be hard – if it were too easy, what fun would that be? I’ll be revealing the identity of the samples shortly.
In the meantime, I should say that the samples rendered from this exercise are nearly identical by virtue of the mechanics involved, and so there should be no hard feelings either way. True, there may be some preferences for one sample over the other, as these comments have expressed, but this is truly pixel-peeping. Whatever one’s preference in these samples, I think most everyone can agree that, in print, any differences between the two would be so close as to be non-existent.
You’re welcome Todd. Well, my guess was based on the slightly different contrast between the 100% crops. Photo A seems to have slightly deeper blacks than photo B, but it’s pretty much a shot in the dark. There’s no real discernible difference between either of the shots.
João,
Yeah, people seem to picking up on the slight difference in contrast, though even then that small clue doesn’t make a clear cut answer. Thanks for your additional thoughts on this!
Without reading any other posts yet: I’d guess that “A” was the native 12800 and “B” was pushed. I think somehow A looks slightly better but I’m not quite sure why. A tiny bit sharper? In the small photos at the top there is practically no difference at all; it’s only noticeable (and I mean barely) in the crops.
I love the results out of that Raw Photo Processor. I don’t do my PP on a Mac though right now so can’t use it. But I’m bookmarking, thanks. Now will go see what other people thought was the original higher ISO.
Hey Maryelle,
Thanks for your thoughts on this exercise. It seems that there is a bit of support for sample A with regard to having an edge in image quality.
RPP has an amazing demosiacing engine – it really does a great job with RAW file and extracting detail. Still, I find that getting images to look natural in terms of color and brightness takes so much work that gains in detail are overshadowed. I do use it for particular landscape images and other shots that benefit from as much detail possible, and which will be printed very large.
[...] of high ISO and push processing, and an introduction to unity gain. Earlier in the week I posed a simple question: Can you tell the difference between an image shot natively at ISO 12800 and one underexposed at [...]
I’ve posted the answer to this little “test,” along with an explanation of the mechanics involved:
Push Processing and Unity Gain
Hey Todd,
I should start by saying I really admire your work, especially having tried concert photography myself. Technically challenging would be a serious understatement!
I was going to say A looks sharper, but I think that’s just the shift in focus as i’m looking at the top half. I can barely notice any other difference. Given that this is at 100%, that’s astounding!
By the way, the link you’ve posted in #33 is to this article, not the conclusion.
Cheers!
Hey Fred,
Thanks very much for the kind words. Yes, concert photography is one of the most demanding types of photography, I think, due to the time limits, constantly changing lighting conditions, and a host of other factors. But I’d say it’s also rewarding for the same reasons and others, too.
Thanks for the heads up about the link in comment #33, should be all fixed now.