Digital Noise - The Truth vs. Hype

[LuisV]

I wrote this in answer to common misconceptions about digital noise. It came to mind because of the recent D300 and D3 releases the inevitable comparisons between Nikon and Canon that comes up when either releases a new body. It's pointless, but always seems to rear it's head. Additionally the topic of noise comes up from time to time, in general, so I came up with the following little series of shots to explain the noise issue a little.

All digital cameras have some level of noise present in all the shots they take. I am not going to get into a huge debate over this, nor do I plan to explain it. It suffices to say that all sensors exhibit some amnount of noise when measured against the capture. It is referred to as Signal to Noise ratio. The better the ratio the less the noise. What I will say is that every single digital camera sold today exhibits noise. I don't care who makes it, I don't care what glass you put in front of it, I don't care if software corrects it out, they all manifest noise in some way. It's inherent in electronics. How they process the noise may vary, but it's there. We can thank marketers for creating this myth that one brand or another can give you clean shots at ISO 3200. It simply is not true. It may be "acceptable" but it's there.

What want this thread to be an excercise to show how that noise is manifest in the images. For the most part (again to simplify this and not make a 5 page discourse) digital noise manifests in the darker areas of the frame. ie The mid to dark shadows of the shots. Also, for the most part, because of the amplification of the signal (light being registerred), as you bump up ISO noise is augmented as well. Again, true of every camera made today. This is exactly the same as it was with film. As you bump up to higher sensitivity film the grain is more pronouced. (Noise and film grain are not the same, but the net result is very similar on the print).

So here's the ticket. I am going to post four shots. You will see there are varying levels of noise in the shots. Your mission, should you choose to accept it, is to tell me which one was shot at what ISO. Well, for arguments sake, which is the highest ISO which is the lowest, ect. Base it on the noise you see, or don't see. I will tell you that the best shot was taken at ISO 100. From there, you have to figure it out.

General notes:

All the shots were taken in studio lit with only a single strobe with a softbox. All the images were made with the same exact camera and lens. (D2X and a 28-70mm f/2.8 lens) All the images were shot in RAW and adjusted so that the third grey square (from left to right) came to approximately 160,160,160 in RGB. This is the standard for the square as listed by Gretag Macbeth. This is also the square used for white balance. The shots are simply labeled A to D. The letter on the full shot matches the crop on the blue square. ie. Crop-A is a crop of Full-A, and so on.

Lastly, please take your best guess. Don't say something different than the previous person simply to get it right if the other is wrong. Also, don't try to d/l the file and read metadata. I am brighter than that and stripped it. . It's not that hard. All the files manifest different levels of noise so it should be easy. Once I have some replies, I'll post which is which.

Now, before a few of you go off on tangents. Cameras all exhibit different capture at differing ISO's. As a result the colors and the blue patches, as a result, may not match. That is not a concern. These shots are simple conversions from the RAW image. I purposely removed my custom curves that would give me the same color on all the patches. So before you go off and say, "hey the colors don't match"..... yes I know and that happens. The only thing adjust in the RAW conversion is the exposure to bring square 3 on the bottom row to 160,160,160. The shots were all sharpened the same way. This sharpening is my capture sharpen to counter the anti-aliasing filter.

Lastly, this is pretty much worst case scenario. So noise here will be most apparent.

Have fun and post those guesses........

Full-A

Full-B

Full-C

Full-D

Crop-A

Crop-B

Crop-C

Crop-D

[Ronald S. Jr.]

Not sure of the order, but I'd bet that C is the highest. Noink cameras, in my experience, have some terribly overpowering noise reduction done in camera, and it kinda ruins the image.

Look at the new D3. Sure, there's very little noise at some of the highest ISO's. However, they almost look like paintings.

[LuisV]

Not sure of the order, but I'd bet that C is the highest. Noink cameras, in my experience, have some terribly overpowering noise reduction done in camera, and it kinda ruins the image.

Look at the new D3. Sure, there's very little noise at some of the highest ISO's. However, they almost look like paintings.

Ron,

C'mon, humor me. I am not asking for exact ISO figures, that's impossible. At least rank them in terms of highest to lowest ISO based on the noise. All four images clearly have different levels.

Nikon's noise reduction is irrelevant here. The images are RAW files. Noise reduction is not applied in camera. That means that these are as they were shot, with the exception of the sharpening (which will augment the noise) and the exposure tweak to get that third square into range.

[Toe Knee"F"]

Crop D...1600
Crop C...100
Crop B...800
Crop A...400

Crop Me...??????? :-)

[WildSwan]

I am so tired of this Noise issue that I will just sit and watch.

I also do not hurt my head so much thinking about the noise... specially if a beautiful model is in front of my camera.

[Cannibalized]

A 400
B 800
C 100
D 1600

C looks really smooshy though. They all look like crap.

I have a canon 20d and even at iso 100 which is as low as it can go it is very noisy in the shadows. it's upsetting to do studio product shots and have crisp lighting but all that noise. but i know the 20d is "old" now even though ive only had it for a year and the sensors are getting better and better so the next camera i get someday should be much improved.

[WildSwan]

Do not despair, Cannibalized... because the noise you see at 100% crop never shows in prints up to 10R size. Even if you print it at 20 x 30, you will not notice it unless you look at the print few inches apart from your nose.

I did a food shot at ISO 1600, available light.. darn, the photo is noisy... but when it got printed, I am surprised that the photo seems to be noise-free.

[LuisV]

Not going to get into debate on the quality of images. These are worst case scenarios and, on top of that, all over-sharpened (based on the type of image) to augment the noise. To assume anything from these is meaningless as it is not even close to what a real world image will look like.

The obvious one was image C. That was shot with proper exposure at ISO 100. This is a clean shot. As it would be on any cameras photo from it's base ISO, whatever that may be.

From there I am now going to make a point that I have made before. Noise is caused as much, if not more, by wrong exposure as it is by high ISO. Put simply, I can make a shot that has more noise shot at ISO 100 than one shot at ISO 800. The proof is right here.

Shot A was shot at ISO 100 just like Shot C. The problem with Shot A is that it was under exposed. The exposure was then raised in the RAW processor to make patch 3 of the bottom row come to standard. This brings the noise way up.

Shot B was shot at ISO 800. Proper Exposure.

Shot C was shot at ISO 100 as mentioned.

Shot D was shot ISO 800 as well. Slightly underexposed and brought up. Not as drastic as the Shot A though.

Each of you picked Shot B to be better than D yet they were both shot at ISO 800. This shows how proper exposure makes a difference.

In my opinion Shot A is worse than both B & D. At the very least, I see it as offensive as the ISO 800 shots.

Overall, the point is that proper exposure is as much, if not more, of an issue than ISO when it comes to noise. I can do this with any camera. It's just the nature of digital sensors. ISO is not the only culprit.

[Cannibalized]

Do not despair, Cannibalized... because the noise you see at 100% crop never shows in prints up to 10R size. Even if you print it at 20 x 30, you will not notice it unless you look at the print few inches apart from your nose.

I did a food shot at ISO 1600, available light.. darn, the photo is noisy... but when it got printed, I am surprised that the photo seems to be noise-free.

i guess ive never tried to print something that looks like poo at 100%. i should, though.

[WildSwan]

And if I may add, when you subject a photo that is properly exposed in a noise remover software, the result is way better than cleaning a badly exposed photo with a noise remover software.