My D300 Images Noisier Than Everyone Else's!

If it matters, "Active D Lighting" is turned on.

Active D-Lighting definitely matters. Turn it off :).

Active D-Lighting does add a bit of noise in the shadow areas, but it's no different from post-processing and boosting the shadows then, which also will add some noise.

What ISO did you shoot this at? If it's 400 or above, there will be some noise. Why is this a problem for you? It's a nice photo.

The EXIF data indicate ISO 800, and Leigh has some noise in the shadow area. The D300 should be able to do a bit better but this doesn't seem to be totally out of line.

Noise is not the only quality measure, nor is it sharpness. Looking at your natural skin colors and the well exposed image, I am tempted to cash out the fortune for the D300.

Of course, D-lighting will lighten the dark areas and thus amplify the noise too. You can remove this with any noise filter, applied on the dark areas only, if you need to. Or with Camera RAW developed twice with different noise settings, and blending the two images. Or you just don't care.

It seems a bit noisy to me too... probably it is due to a slight overexposure, thought. Check my attached image: same settings as you, except for the NR (which I think it must have a little effect at 800 ASA), straight RAW file converted to .jpg. D-Lightning set to "normal".

My scene were really dark except for the Coke dispenser room; I exposed for that dispenser. Check that the dispenser is bright black, it shows a little noise, but the also very dark corners of the frame show a little higher noise... this areas are overexposed thought. I`m not an expert here, perhaps others could corroborate (or not) my words. (Please no comment allowed about the image, I know it`s one of my worst ones!) ;P

I`ve found D-Lightning is very convenient on most shots, but annoying on others, more thoughtful shots, where light can be under control.

I doubt that your D70 would have done nearly as good as your D300 with this shot. Did you shoot RAW or JPG? Have you tried opening and processing the file with NX rather than an Adobe product?

My guess is that the 'problem', if there truly is one, lies with your post processing (either the program or your settings).

Or it was underexposed to start with and was boosted automatically by the raw editor.

If you're using Photoshop, what settings were set automatically in Camera RAW - specifically the Exposure slider?

Leigh, that sample photo looks much better than my D2H at 800 out of the camera. While the luminance noise ("grain") is comparable the D300 has far less chroma noise (splotchy discoloration). It's also far better than anything you could do with 800 color film.

Not saying there's no room for improvement, but if you want truly noise-free high ISO photos you'll just need to keep refining your technique and possibly use a dedicated noise reduction software like Noise Ninja (there are others, I'm just familiar with that one).

Thanks for the responses. I'm not terribly unhappy the the noise levels; it's just that others are raving about the amazing lack of noise, and I don't seem to be getting those kind of results. If you search these forums for "D300" & "noise," you'll find lots of examples of shots taken at ISO 1600 and above with virtually no noise at all. My shot shown here is only ISO 800 and is quite grainy. Does anyone set their in-camera noise reduction to "High"? I don't mind the noise and can deal with it with Photoshop or its plug-ins, but I just don't like the nagging feeling that my shots are not supposed to be this grainy with the D300.

Regarding the question about the exposure, it was shot in raw (NEF) and I boosted the exposure by 0.85. Luminance smoothing and color noise reduction are both set to 25.

Leigh, if you got a free copy of Nikon Capture with your D 300, try loading it and processing your NEF image with it and see if the noise is reduced. Joe Smith

I think there may be something amiss since the noise in the brown tones at the back looks similar to the brown tones in an untweaked camera JPEG from my D40 at ISO 900 (I used auto ISO) and I'd expect a D300 to be better.

I was actually trying to match the noise/sharpness myself on a RAW image and failing, although this is only with GIMP so don't read to much into it.

I was going to say try an in camera JPEG of something similar and compare but Joseph's idea of trying Nikon Capture may be equivalent since it may use the same algorithm as the in camera procedure.

It's a really good picture anyway.

If you can bear to look at a photograph of a pile of untidy clothes, here is a 1/4 of the D40 frame:

Ditto Joseph's suggestion to try Nikon's own software for NEFs. When I got my D2H I tried every software package available that handled NEFs. Nikon's own software consistently produced the best results. Not the best software, necessarily, and some folks continue to report it's a bit sluggish even on fast computers. But for maximum image quality it's probably the way to go.

Also, Leigh, you mentioned boosting the exposure slightly. If that's the case, the original photo may have been slightly underexposed. This will always contribute to noise, even at lower ISOs.

Occasionally I've been forced to use my D2H at the Hi-1 and Hi-2 settings (ISO equivalent 3200 and 6400, respectively). It can actually produce usable photos, especially if I convert to grayscale, with noise resembling grain from Ilford Delta 3200 at 3200-6400 (popcorn fluffy). But it is completely unforgiving of underexposure. Even the slightest underexposure produces banding. I can deal with the noise but banding produces a useless photo. So I have to discipline myself to remember that with digital, ISO 800 means 800, not 1000; ditto 400, 1600, 6400. No fudge factor.

no, your photo is not noisier than everyone else's. it really depends on the circumstances of the shot. this bird was in the wrong place at the wrong time -- the sun should've been at my back, not his. but he was BIG so i stopped the car to do a photograph. when viewed later at 100 percent, it was quite noisy. ISO at 640, by the way, and i've turned off active d-lighting in captureNX. you can shoot at higher ISO and get acceptable results with a D300, but you can still get noise.

the image to me is lighted ok, now. but if you boosted the exposure by .85 stop then it was underexposed by almost a full stop when shot. underexposing is guerenteed to produce noise. when you shoot you are always trying to maxamize the S/N ratio. the way to do that is cram as much signal(image data) into the shot as you can, without blowing the highlights.

the way to get the signal amount high is to expose to the right. that is on the histogram, i think the d300 has 4way histogram, put the histo line as close as you can to the right wall without actually touching the wall. the histo line should touch the floor of the histogram just before it reaches the right wall. this maximizes the signal and decreases the noise to lowest amount possible. if the resulting image is too bright then simply hit auto levels in photoshop or pe5-6 to get the light at the normal amount. the light amount is then reduced to a normal level and the image noise is reduced along with it. but if you underexpose, as you did, when the image brightness is increased to normal, the noise amount is increased as well. you are far better off shooting to the right and increasing the signal and decreasing the noise.

I believe there is another issue in comparison of the D70 and the D300 capabilities: You can take advantage of the D300's higher resolution. Compressing the photo will reduce noise. If you shoot "Large" on D300, which is 4288x2848 pixels, and reduces the photo to ca 3000x2000, which is max resolution on the D70, then you can compare. I believe you will find D300 far superior. And what is already said about avoiding underexposure is very significant. I attach an example of a full resolution crop from an ISO 720 JPEG shot from my D300 (no post prosessing). As you can see, this is not noise-less. But when reduced to PC screen resolution (72 ppi), it will look quite good.

... but if you underexpose, as you did, when the image brightness is increased to normal, the noise amount is increased as well.

It could be similar to overdevelopment of an underexposed film. It produce similar results on grain. How could be called this process on digital? Boosting? Oversaturation?

Well if that's the case, then I'm sure happy I have a 40D rather than a 300D. Even at ISO 1600, I've never had an image even remotely that noisy.

Jon,

I wondered the same as you about the number of pixels making the D300 example look bad when compared with my D40 one so I downsampled Leigh's original to 2904x2205 and compared again.

I think my D40 image actually looks less noisy both before and after the downsample. The ratio is sqrt(2) which should be the correct factor since the D300 has double the pixels. I used GIMP for the downsample. I don't know if Photoshop or something else might be different.

Darn - I was going to install CaptureNX and see if it handled noise better than Adobe Camera Raw, but apparently my D300 only came with ViewNX, which does not appear to be an editing program.

Try in camera JPEG. I think that is what Jon's portrait uses and it looks much less noisy.

I'm interested in this since I'd quite like the D300 in D80 clothing when it eventually arrives!

I think that there are several issues going on with this image. The first is the use of ACR. Frankly, I believe that Adobe still has some catching up to do in regards to the files from the D300. I would suggest that you try using Capture NX on your image and see the difference for yourself.

BTW, ACR will ignore the following settings during the conversion. First, the AD-L will be ignored, I believe that is a tag that triggers NX to process the image using the setting in the camera. So, the good news is that your noise was not due to the use of AD-L. The second issue is that any Noise Reduction you have set in the camera is also done by NX during the conversion. That means that ACR imported your image without any noise reduction. The third issue is the underexposure that you had to correct for. Anytime you underexpose an image using a digital camera, increased noise is a sure result. I also have to ask if you did any sharpening to your image. If you did, this will drastically increase the noise of the image if it's done prior to applying some noise reduction.

My advice is that you either consider using NX if you want to keep on shooting RAW, or consider taking pictures like this using JPG.

I will let you in on a little secret here, the JPG "engine" in the D300 is simply superb. It is so good that about the only time that I now shoot RAW is if the lighting is really funky. I'll grant that shooting JPG doesn't provide the "cushion" that shooting RAW permits but, if your careful about getting everything correct, you can shoot JPG and produce images that are capable of producing results that are essentially identical to what can be done shooting RAW. So, you may find it worthwhile spending some time experimenting with shooting JPG and learn how the camera responds well enough that you can shoot JPG and not spend all that time converting and post processing images shot in RAW. As for my tips for shooting JPG, basically I set the camera so that it "tinkers" with the image the least amount possible. Specifically, I leave the High ISO Noise Reduction set to OFF, as well as the AD-L. These are items that can be corrected in NX with a much higher degree of control than the in camera settings. I also have my Sharpening in the camera set to level 2. Which is fairly low but the images respond so well to Unsharp Mask that I just don't see any reason to risk the increase in noise that a higher setting would cause. I am also very careful about exposure and white balance because shooting JPG will not permit the flexibility for correction that shooting RAW permits. My point here is that if your very careful, you can shoot JPG without any deficit.

OK. Well, apparently CaptureNX has a free demo, so I downloaded and installed. I worked up the same photo to try to duplicate the settings and compare noise. CaptureNX and ACR are apples and oranges, as you can see from the different white balance, but I will say the noise is significantly reduced with CaptureNX. However, the whole picture is softer, too, so I'm not sure I like it better. I'll post it again later after applying some Pshop noise reduction...

And now, the same picture but processed in Photoshop with Adobe Camera Raw with applied "Reduce Noise" filter. Again, it's hard to compare noise levels because exposure and white balance are different, but I think the lesson for me is that ACR is fine for processing my raw files (holy smokes- CaptureNX is SLOW!!!) and that the key to noise is exposure. This shot is definitely underexposed, and that seems to have been the key problem.

Leigh, I'm seeing very good detail preservation in the Capture NX version with much less luminance noise. Looking at individual strands of hair, the detail is very well preserved.

What I am seeing is very slightly less micro-contrast, which gives the impression of less detail in certain areas, such as the fabric of the sofa. But true detail is well preserved.

WB and color tweaks are always a challenge. Between Nikon, Adobe, Bibble and various other raw converters, I can't say I ever found one that was consistently better at rendering skin tones to my liking. It always depended on the color temperature of the lighting, differences in skin tones and voodoo.

May I remind everything not to post those full-size, 3, 4 mega byte image files? Those large files are very demanding to our bandwidth, even though storage space is cheap now.

But I would like to see this discussion come to a natural conclusion in another day or two, but I'll have no choice to delete this entire thread in order to remove all the attached images.

In the future, please only post relative samll JPEG files. If you need to show details so that we can all see noise and sharpness, please just crop a small section from your entire image.

Thanks.

Oops... In my post above I meant that my Canon 40D is never even remotely as noisy as that picture... even at 1600. If that's what kind of noise can be expected out of the D300, I think i'll stick with my Canon thanks

I think D-lighting will have had a significant effect on the original poster's photo - it is quite likely that the camera underexposed to preserve the highlights in the window, and boosted the signal in the shadows in the rest of the image (hence the noise).

To avoid this effect, I would suggest metering on the subjects face using spot or center-weighted metering, or using a speedlight to provide fill light.

Can you see the indicated exposure go down when you switch AD-L on with a high contrast scene?

If that's the case then presumably it reduces exposure for contrasty scenes just enough to stop the highlights blowing and then recovers shadow detail by applying a non-linear curve - much like one would do oneself with a little tweaking after looking at the histogram.

First I would ask question, was that necessory to take a shutter speed 1/20 (as exif) in so high ISO, I think you could have put much faster in the situation....

Noise always come in any medium with high ISOs, if you want cleaner images, than buy a fully autometic, big sensor point & shoot digital camera....

Some friends here already suggested some settings like D-lightings, or others... Try those

it should be mentioned to all the canon 40d users who are saying that the 40d is less noisy. not necessarily true. ANY camera can be made to appear noisy in the images it takes if the shots are underexposed by a stop or more. and this is exactly what happened in the subject of this thread's photo.

in any digital photography the user must expose to the right to maximize the signal and keep the noise at a minamum. this was not done by the writer of the thread.

by the way, i own neither canon or nikon.

It's the Active-D which is giving the noise in the upper left shadow. Jose's Coke machine with "D" off clearly shows a difference

Hi! I think there are several misconceptions around about noise. My D70s at iSO800 looks definitively better, but as mentioned above, this has to do with sensor size and pixel size. Please correct me I were wrong, but D70s & D300 are both DX format, so they should have the same sensor size. This means that pixel size on D300 is smaller (more pixels packed in the same area), and when equally exposed, D300 pixels capture less charge per impinging photon, and obviously the signal-to-noise ratio is worse. However, downsampling (=low-pass filtering or blurring + subsampling the image array) may decrease noise levels (because it is averaged out if noise statistics is spatially uncorrelated), thus at the same resolution 3000 x 2000 D70s & D300 should appear similar (take a look at http://www.6mpixel.org/en/ for a nice illustration, although they discuss the issue with compacts).

Next we have noise reduction algorithms, in-camera and post-processing. I do not know much about the camera-intern algorithms, but noise-ninja, noiseware, or neatimage use multi-scale wavelet analysis (or a similar oriented bandpass decomposition) and then analyse wavelet coefficients more or less successful to get rid of noise (there are a lot of articles out there in the IEEE Transactions on PAMI or Image Processing for the interested). Since noise perceptually affects most the fine details in an image, any noise-processing algorithm has to make a sort of decision whether a high spatial frequency content at some location was caused by noise or is due to image structure. The more noise the worse the less fine detail can be conserved. Camera-intern noise suppression in D300 has most likely been enhanced with respect to D70s, so people claim that D300 images look better. However, when viewed at high resolution, you should perceive a loss of fine detail (typically boundaries look jagged), although uniform structures look indeed smoother. Physically, D300 has more noise due to its smaller pixel size.

Applying D-Lighting enhances luminance of small-intensity regions, but small intensity regions are at the same time more prone to noise. So noise (or what was left by the camera-intern noise suppression) is enhanced. Somebody suggested a good strategy to save the RAW-files and use post-processing algorithms for noise reduction on 16-bit TIFFS. If you are good at programming and maths, you may then re-implement a state-of-the-art method proposed in some of the articles (most authors provide code, usually in Matlab, along with preprints at their websites). Recently published methods should outperform what you can buy, but nevertheless I confess to also use some professional tool because Matlab is not as fast and comfortable (unless you reprogram it in C etc). Recently, researchers found that combining noise-reduction and demosaicking is superior to first demosaicking and then reducing noise. So, future promises exciting new algorithms.

Comments welcome - Matthias

Matthias,

Thanks for that it was very interesting. I tried to email you directly but the photo.net email sent me a bounce so perhaps they don't have your latest address.

Have you tried "GREYCstoration"

http://cimg.sourceforge.net/greycstoration/index.shtml

It's free and open source and after just a little time with it, it looks as if it's going to do at least as well as the algorithms built in to the D40. Hopefully better with more time to experiment.

Dear Richard,

Thank you very much for providing the reference, I did not know it! But this will change in the next few days, I will have a look at the papers and the method :-) Looks promising, and I am interested how the PDE-based regularization compares to state-of-the-art wavelet approaches.

Thanks for your input!

Matthias