High ISO imagery

[Mike Umscheid]

Which image of the 2 below do you think appears less noisier to you... the one you think would be easier to take care of with just routine noise reduction through Neat Image? (or whatever your favorite noise reduction software is). I basically took a picture of one of my 16x20 prints for this test.

Each image was taken with the same lens, but a different camera, and care was taken to do this test with the same exact exposure and distance to subject throughout. The two test cameras were the Nikon D200 and the Nikon D3. Both images were also given the same amount of Curves adjustment to open up shadow tones, which also brings out the noise more. These are untouched embedded JPEG's from the RAW files (this can be done using the DCRaw program). Each camera has the same High ISO noise reduction setting (set at Standard). Reply with your thoughts. I'll post tomorrow the details of each image. By the way, I was very, very pleased with the results of this test

Image #1:

Image #2:

 

[Chris Allington]

Number 2 looks less noisy and doesn't have some of the off color artifacts like the first image does.

 

[Mike Peregrine]

The second one does look like it has less noise, but this is a little tricky. The first one actually looks sharper in the detail to me and has better contrast, which makes me think that it might get better overall results during the clean-up. The second one is a little soft already, so running Neat Image on it will probably make it softer and lighter still. There are different types of noise caused by things other than ISO in photos ... and some is easier to correct than others. I'm going to go against my first instinct and say the top one will turn out better.

 

[Dustin Wilcox]

I agree the Second one has less noise but the first is much sharper and show much better contrast, there are some tiny parts of the lightning branches that show up on pic #1 but cease to exist on #2...

 

[Mike Umscheid]

Thanks Dustin, Mike, and Chris. I tried to use the curve adjustment (the only processing that was done to both) to make the tones appear similar. Some of the things that Mike mentions may have to do with some of the in-camera processing settings I have on the D200 for things like contrast and saturation boost. This exercise is something I wanted to do to compare noise handling between these two cameras, so that I know how flexible I can be with ISO in the field with the D3. Also, I would almost never move deep shadows this much in actual processing. The funky blotching has to do with light reflection off the glass in front of the print, but that's moot.

 

[Mike Umscheid]

Here are the details of the images (both shot with Nikkor 14-24mm f/2.8):

Image 1 -- Nikon D200, 1.0s, f/5.6, 14mm (21mm equiv.), ISO 640
Image 2 -- Nikon D3, 1/10s, f/5.6, 14mm, ISO 2500

That's a solid 2 stops (or more) increase in ISO performance. The reason I chose ISO 640 was that is about is high of an ISO I felt comfortable going on my D200 (and D70 which I'm pretty much retiring now). I hated to shoot ISO 800 and only once in a blue moon shot ISO 1000. The Nikon D3 performs *better* at ISO 2500 than my Nikon D200 at ISO 640! I was hoping for at least a 1 stop improvement -- am really loving 2 or more! The thing about the Nikon D3 noise is the fact that it is very fine... not nearly as coarse as is seen on a majority of other dSLRs. This would usually mean an easier time for Neat Image or Noise Ninja. Other photographers have noted the same performance on the D3 High ISO performance:

http://ishootshows.com/2008/01/07/impressions-on-the-nikon-d3-high-iso-performance/

http://www.kenrockwell.com/nikon/d3/vs-5d-180mm.htm

This kind of performance is fantastic for photojournalists, wedding photogs, concert photogs, etc.. but this does open up some fantastic opportunities for (semi) professional storm photographers, since our subjects are not stationary (i.e. landscape photographers) and the best storms tend to occur around the dusk hours. Also, there are many instances when there is just no time to set up a tripod, too.

 

[Mike Peregrine]

That's pretty amazing. Wouldn't have believed it. You'll be getting some amazing storm images with that kind of ISO power. Have fun and thanks for sharing the test -

 

[Bill Hark]

Mike, thanks for posting the images. It's nice to see tests of equipment under real storm conditions. I am amazed that you can get that decent of an image from iso 2500.

Bill Hark

 

[Ryan McGinnis]

The first one is definitely worse off. The new Nikon bodies are finally surpassing the Canon models in quality; apparently, ISO sensitivity is the next frontier. I mean, photographers are willing to pay $3000 for that extra stop on their telephotos -- you're getting two free stops with any lens with all the new Nikon models. Pretty amazing stuff. I can't wait for the day when we can take available light snapshots in situations that require a tripod right now.

 

[Darren Addy]

Interesting experiment. I don't mean to quibble, but the exposures are not exactly equivalent (based upon the ISOs), which they should be if you are going to judge apples to apples.

Image 1 -- Nikon D200, 1.0s, f/5.6, 14mm (21mm equiv.), ISO 640
Image 2 -- Nikon D3, 1/10s, f/5.6, 14mm, ISO 2500

Since the aperture remains the same, one stop improvement (effective ISO of 1280) would be a shutter speed of .5s. Another stop improvement (effective ISO of 2560) should be a .25s shutter speed. However...
Image 2 is more than another stop under-exposed than that at .10s. While that is even more impressive than a 2 stop claim, it makes me wonder what the differences look like if the exposure is taken at what the ISO really indicates. In effect shot 2 is both taken at a higher ISO and more underexposed (two variables have changed, not one).

This raises questions in my mind: Does this indicate a problem with the camera's interpolation of exposure at different ISOs or that the camera is not giving you the actual ISO it says it is (or was this a manual exposure setting at the different ISOs). Or does it indicate that post processing can do more with an underexposed RAW? Something doesn't add up completely somewhere.

I'd also rather see the test on a real subject than on a photo of a print (which sort of messes with one's mind).
: )

I love the experimentation. It's just that I would love to refine the experiment a little more and isolate the variables to make sure the proper conclusions are being drawn from the results.

 

[Mike Umscheid]

Your math isn't fuzzy Darren You are quite correct. This was indeed a very poor man's experiment, certainly not scientific by any means. The full frame of the 2nd image is about a third to a half stop underexposed of the first (when comparing histograms of each of the full frames -- these examples are tight crops), which is a little more impressive as you say. I can't totally explain the reason that the exposure readings don't match up perfectly, but that's another topic. I was shooting in full manual, btw. It may have to do with the full frame sensor vs. the smaller DX sensor? Maybe I should research this. At any rate, it's enough for me to concluded that, yes, I can feel comfortable shooting this camera at insane (tm Hollingshead) ISOs that I had never previously dreamed of! I totally agree with Ryan, as imaging sensor technology improves, it will continue to push photography into a whole new realm. With these glowing reviews of the Nikon D3, I'm sure Canon has a big response up there sleeves coming up. That's how this competition works, and it's great for the consumers! Next test: storms in low light w/o a tripod at 3200 ISO

 

[Mike Peregrine]

If Canon can do the same thing, there will be more than one storm chaser (this one included) who will be jumping on board. It would open up so many more possibilities to be able to shoot in very low light without worrying about noise or losing quality (what I wouldn't have given for this under the Glen Elder storm base!!!). I can't wait to see what you come up with under a dark storm at 3200 ISO. Go find one right now please.