Terrible Noise/Pixelation With D300 - Active-D Lighting

Discussion in 'Nikon' started by whoz_the_man_huh, Mar 17, 2009.

  1. Hi,

    Here is the full-sized JPG, converted from 14-bit RAW by Capture NX 2.1.0:


    Full res photo


    As you can see after zooming in, water reflections and darker trees show an unacceptable amount of noise and pixel smearing.

    I used the Nikkor 16-85mm VRII lens with the following settings:


    F11
    ISO-200
    1/50 sec


    Can anyone tell me why this is happening? My photo guru friends are at a loss, although the RAW converter and resolving power of the lens have been brought up as suspects.

    Thanks,

    Calvin
     
  2. Perhaps you could post a crop that shows exactly what you are referring to. For me, that looks just fine at 100%.
     
  3. Joel, what if you were to go to 200%?

    Looking at the reflection of the dark green trees near the center, do you see pixels and black dots rather than a clean and clear water surface?
     
  4. Flat, gray lighting and no use of the usual sharpening and curve adjustments in post will certainly make for a drab image, as shown. But I'm not seeing pixelation or noise - just a somewhat awkward exposure and some softness (which could be any number of things). There's also a green smudge of lens flare near the bottom center-left (was the hood on the lens? if not, that might account for some of the lack of contrast).

    Is there any chance that you're looking at the image on a display that's producing some funky 6-bit banding effects or other artifacts?
     
  5. Calvin, I'm not sure what you are referring to as "noise". Your image looks very clean in the water, sky and shadows. Maybe it would be helpful for you to describe your specific concerns about this image.
    Dick
     
  6. Matt, can you please zoom in on the reflection of the dark green trees at center-left? That is where the strange noise is most apparent.
    I did have my lens hood on although I'm not sure I know which smudge of lens flare you're referring to. And I've confirmed the issue is visible on several different monitors.
     
  7. Your camera is set to boost saturation, i doubt that is a problem, but if you use a raw converter, why bother with the in camera saturation setting?
    Since you took the shot in manual mode was it way under exposed before you put it through the RAW conversion process? If so, theres you answer.
     
  8. Richard, it's the water reflections that seem filled with noise when you zoom in close. I don't see clean green and brown but rather pixelated patches that are sprinkled with black dots.
     
  9. Thanks Galen, but the photo never appeared underexposed.
     
  10. Just for those who don't want to have to open that whole file, here's a thread-friendly 100% crop from the area that (I believe) Calvin's referring to.
    00Smls-117079584.jpg
     
  11. Wow. Thanks, Matt! That's precisely what I'm referring to.
     
  12. Calvin, I switched from my small laptop to a 24 inch iMac and I now see what you mean. It is in the trees, the dark reflections and the wood of the dock on the right. I can't explain this, but maybe someone here will.
    Dick
     
  13. To me, it looks like typical chroma noise that results from trying to boost shadow detail in s scene dominated by a bright area (the sky, in this case). I'm guessing that this is a result of having the Active D-Lighting set to "High" (which I see, in the EXIF data, is the case).
     
  14. I see the stuff that the OP is talking about. This is possibly a combination of pixel peeping and maybe a bit underexposure in the dark areas of a high dynamic range photo. (I'm not somewhere where I can really look at the RGB curves right now.)
    The noise doesn't really look all that unusual to me. Point the camera at a clear blue sky and you'll probably see something similar. But if you post the RAW file somewhere it can be downloaded it, I'd be happy to check it out for you.
    I never bother to shoot 14-bits because I don't see any difference with 12-bits for my applications.
     
  15. Thanks, guys! I'll try disabling ADL tonight and let you know.
     
  16. just curious, calvin: how long have you had your D300?
     
  17. I agree that active D lighting being set to high isn't a good idea. When shooting raw I leave D lighting off as it can always be added later in Capture NX or NX2.
     
  18. Hopefully something derogatory is not headed my way, William.
    I'm a true newb. I've had the D300, and coincidentally been off full-auto mode, for about two and a half weeks.
     
  19. Was it underexposed to begin with?
     
  20. It wasn't, Ryan.
     
  21. If you zoom in beyond 100% magnification you will begin to see pixels and artifacts with any photo. This is normal. It's not like TV or the movies where they can zoom in effortlessly and magnify photos while retaining full detail. That's fiction.
    Photos can be resampled to higher resolution, but this is a different process from increasing onscreen magnification beyond 100%.
     
  22. Don't worry, Lex. I don't expect a fictional level of detail.
     
  23. D-Lighting and Active D-Lighting aren't the same thing. You can indeed apply classic D-Lighting after the fact in NX. But Active D-Lighting essentially plays with the nature of the exposure at the time it's taken. It's the camera's way of altering the exposure to avoid clipping. When you put it on "High," it gets very aggressive, and tries to compress more of the prospective image's info into the middle part of the curve. Something always has to give... and when you essentially increase gain on those shadowy areas, some chroma noise always shows up.

    However: I guarantee that if you applied a wee bit of NR in post, and made a print (say, an 8x12) you'd never even see it. As Lex points out - zooming in past 100% doesn't shows you anything meaningful, but can get you worried about nothing of consequence.
     
  24. Here's an example of the differences between onscreen magnification and resampling. All photos straight from my D2H at highest ISO, 6400, which is extremely noisy. I left the flaws intact to show the distinct effects of image degradation from both magnification and resampling.
    The second and third photos were each resampled *downward* to 25% of original size.
    #2 was then magnified using the usual onscreen magnification tools of the pixel peeper until it reach the same 500x500 pixel size as the original. This is an example of what you will also see on the digital camera's built in LCD review screen at maximum magnification. The jaggies and pixel artifacts are clearly visible.
    A screen capture was made. #3 was resampled downward to 25% of the original size using the Lanczos algorithm; then resampled upward back to 500x500. While the image is obviously badly degraded, it does not show jaggies, stairstepping and sharp edged pixel artifacts because the resampling algorithm does not merely magnify the existing pixels.
    00SmqZ-117111584.jpg
     
  25. I must admit I never bother much with photos I have not photoshopped at least to (a) reduce noise (b) optiimise contrast / tone (c) optimise color and (d) sharpen. Often I will also then use the local contrast filter which greatly aids image clarity and apparent sharpness. Then I feel I can look to see how good the image is.
    So often, really ordinary if not cruddy images have come up tops after this basic processing that I believe no digital camera is really capable of turning out usable shots straight out of the camera. As a result every photo of mine that I want to keep gets this treatment to get it to the point where its decent. (I am presently mainly using a D200 and am pretty happy with performance although photos straight out of the camera need this work over - just as with every other camera I have owned or tried.
    BTW I especially find that many cameras struggle with images like this one - with lots of fine detail in branches and twigs. Of course to add insult to injury somethimes lenses will also produce CA which is noticeable in this type of image - although I think yours have avoided this fault. Honestly though I think I would not worry until I had processed the image properly and then only I did not like the result in successive images.
    In your specific photo there is a fair bit of blurring (which I might be inclined to put down to focus error or camera shake (one one fiftieth of a second is not all THAT fast) My camera will also sometimes produce surprisingly large amounts of noise under some conditions but not often enough to worry me. it just happens. Most times it does not.
     
  26. So, now that I'm home and have opened your image in CNX2 (rather than within Firefox on the crappy Dell LCD at work) I can see what you are speaking of.
    If I take that image and apply noise reduction at 7, 5, Better quality, opacity: Luminance & Chrominance @ 30, 100 the noise basically goes away.
    I would suggest that ADL at high is not that bad of a problem, since you are still having some rather bad clipping, and it would be somewhat worse if you hadn't used it. I wouldn't have used it, but would have adjusted the exposure to just barely not clip using the histograms on the LCD. In order to recover the image to my own personal tastes with a minimum of effort, try Auto Levels, advanced set to something between 20 - 40 for contrast adjustment and 0 - 25 for the color cast adjustment.
    Finally, I'd selectively sharpen the final image with some USM and the Selection Control Point to only sharpen the areas I want sharpened (or conversely, to deselect the areas I didn't want sharpened. I usually have an upper limit for those of 20, 5, and not less than 5 for the threshold.
    Cheers.
     
  27. There's minimal CA (chromatic aberration) because it's shot at f/11.
     
  28. I see the same problem (apparently randomly) on my D200 raw images, and it's not the same thing that Lex's cat shows. Usually it occurs when I've been taking a lot of shots in a fairly short time and when the scene has too much dynamic range or I've been switching ISO back and forth (it happens at ISO 100). It may be a banding problem (which some D200 are supposed to have due to a sensor problem) or something else. Either way, I didn't think it would also show up on a D300. I never see this for well/over exposed regions (I usually expose-to-the-right).
     
  29. Hi all,

    Thanks for sharing your vast knowledge.

    I went back to the park after work and it looks like Matt hit the bullseye. Disabling ADL removed all noise and artifacts:


    Full res photo


    Hopefully this helps another beginner in the future.

    Calvin


    P.S.: This was extremely timely assistance as I am going on vacation this Saturday. Thanks so much again!
     
  30. Great news, Calvin. For what it's worth: if you have time, try settling ADL down to "normal," too, to see the difference between that and "high," since there definitely is a difference.

    I've been slowly dragging ADL into my thought process while out shooting the sort of stuff I shoot, and when it's not trying too hard, it can be a huge asset. For example, I sometimes shoot scenes on poorly lit gray days where I have to crank up the ISO to get good enough shutter speed to freeze action. When there are contrasty elements in the scene, ADL seems to err on the side of ever so slightly under exposing - and teasing up shadow detail in post will always bring up a little chroma noise. But I'm preseving highlight details that I otherwise would have lost, and these things are always a balancing act.
     
  31. Calvin, thanks for posting the second shot. And, Matt, thanks for pointing out that it was related to the high ADL setting. This was very helpful.
    Dick
     
  32. Actually this is a overexposed capture, may be at the shooting time or RAw conversion time, but it is. So that the highlights have blown out. I think this is not the correct exposure for this kind of scene, in your full resolution posted image, I can clearly see that you tried to get more details in shadowed areas, so it happened but it is not as much of noise according to the lighting.
    I just applyid the Auto Color and after that Auto Levels in PS, and now I don't find that "noise" which you say unacceptable...
    00SnLv-117375584.jpg
     
  33. Matt is right on the money....AND, we only got one "you should have used a Canon" comment. Things are looking up!
    ADL can be very aggressive in such a situation. What you actually did was solve the problem by taking the shot naturally.
     
  34. Thanks, Matt. I'll tiptoe toward ADL again for experimentation purposes although I'm very wary of it at the moment.
     
  35. I don't think the resolution of the problem had anything to do with Canon, mansito.
     
  36. stb

    stb

    Am I being insensitive if I point out that, in my opinion, the OP noticed the so-called problem because there is nothing else to notice in that picture?
    That maybe, just maybe, in a picture with a captivating subject, none of this would matter?
     
  37. Stephane, you're being truthful, not insensitive.
    However, as far as none of this mattering in a photo with a captivating subject, suppose you have two versions of such a photo. One has the pixelation issue and the other doesn't. Which version would you keep?
     
  38. stb

    stb

    Calvin, that situation simply never happens.
     
  39. If someone was to take a captivating landscape shot, notice it has this pixelation issue, and then reshoot the same shot without the problem, I would favor the second version.
     
  40. stb

    stb

    Reshooting the same picture is a myth. You can go again at the same place, you might a second great shot, but it will be another one. If they are both great, you'll keep both.
     
  41. It just comes down to what we individually consider great. The kind of pixelation and noise shown in the original photo I posted precludes greatness in my opinion.
    Having the second photo, I'd toss the first without a thought whereas you wouldn't find anything displeasing about it.
     
  42. stb

    stb

    Oh come on! Just print the damn picture and see if you still notice anything. I bet you don't have a big enough printer to have a chance to notice it on print. At some stage you even told someone to zoom in at 200% to see it. How meaningful is that? Display it as big as you screen allows to see it completely and reflect on that "terrible pixelation" (your words).
     
  43. Hmm? Why would what a newbie throws away bother you?
     
  44. Stephane's point, Calvin, is that pixel peeping will make you crazy, and for no reason. When you zoom to 100% on your typical display, that's like looking at a print that's several feet across. By the time you render those prints at a normal-ish size (say, a 12x18 - which is a pretty big print), you simply won't be able to see that stuff. Getting distracted by it can lead you away from worrying about the larger composition, narration, and technique issues that actually make or break an image in real life, practical terms.

    Also, just to help future communication: noise is a very, very different thing than "pixelation."

    Think of pixelation as situation when an image that has higher resolution and should be showing more subtle changes between two or more adjacent pixels instead shows them lumped together into adjacent pixels that share the same value. Your image has the same number of pixels not matter how much the detail is wrecked by compression artifacts. But it's compression artifacts that cause a handful of subtly different adjacent pixels to all wind up showing the same color/tone, and thus looking like one, larger, chunkier pixel. Hence pixelation. Of course, this can also happen when you take a low-resolution image and up-size it to include more pixels - but that's not what's at hand, here.

    The only thing on the table in this case is noise. And the only thing that really mattered was chroma noise. Depending on the software you're using, you'd be amazed at how well that can be brought under control. But before you do anything more to an image, spend the $1.00 a few times to actually produce prints. It will completely alter your thinking on the subject.
     
  45. Matt, thanks for educating me. You rock.
    I do large prints. One of the reasons I upgraded from the D40 was to get good quality beyond the size of my last print at 16 x 20. That is, I will go to 20 x 30 once I have a shot I really like. Why? The reason is I'm no expert, but rather just a guy who likes big prints of his wife. And of course the photos I develop of my wife and I together are not even taken by me.
    As you suggested, I'm trying my best to develop my poor technique. Your fixing the noise issue for me once and for all has no downside, as far as I can see. Thanks a million again.
     
  46. The key to the whole thing (since, at 20x30, you're definitely at your camera's boundaries) is proper exposure. You definitely want to keep the ol' "expose to the right" (look it up!) in mind, so that the details in the darker tones to have to suffer the noisy degradation of being pumped up after the fact. The histogram display can be very helpful, in the field. But more than that, it's a matter of learning when and how to use which metering technique. That's the big one. Your camera definitely can't read your mind, and despite Nikon's brilliant engineering, it can't know what's most important to you in a given scene. It will seem contrary to most newer users' thoughts, but the sooner you get away from having the camera's metering system do all the work, the better.

    Don't get me wrong - I'm an Aperture Priority kinda guy. I let the camera figure it out all the time. But I help it out by deciding when to use spot, vs. matrix metering... or when to compensate by a stop or so when I know that a bright sky is going to torture my results. That just takes lots and lots of shooting. The good news is that with a D300, you have an incredible tool to work with, and nowhere to go but up in terms of results. Before you know it, all of this stuff will take a distant mental back seat to what really matters: looking through the view finder, and asking yourself, "Why am I about to make this photograph? What can I do with my composition, or with where I'm standing, or with with direction the light is shingin, to make it ten times better?"
     
  47. Have you considered writing a philosophy of photography book, Matt? I'd buy 10 copies.
     
  48. Well, that might sound like a good idea at first, Calvin, but I suspect that the world gets enough of my ramblings and mutterings already. Now, I'm going to put down the keyboard, and pick up the camera, and practice a bit of what I'm preaching. Note to self: take no silly pictures for the next 24 hours. Only Important Photographs. Oh, look! A dog! I'm hopeless.
     
  49. Calvin, you might be very interested in this discussion by Emil Martinec (physics prof from U of Chicago, who is an expert in digital imaging and string theory). As he explains, shadow noise is going to be greater at ISO 100 or 200 and near optimal at ISO 1600. This is a real surprise to just about everybody! He proposes a bi-amplified analog front end that processes low level signals at ISO 1600 and highlights at ISO 200 and demonstrates the results with his "teddy bear and lamp" test. This is clearly the kind of finding that would benefit the images you are trying to capture. You can shoot two exposures of the same scene at these different ISO levesl and blend them. But either way, you can see how the expectations of getting clean shadow detail at low ISOs is just a myth.
    Sensor DR vs Camera DR: Open Talk Forum: Digital Photography Review
     
  50. Thanks for the interesting link, Luke.
     

Share This Page