D700 color bands in sky.. revisited

Discussion in 'Nikon' started by ian_watt|3, Nov 3, 2010.

  1. I've had this problem with my camera since day 1. Nikon have had the camera in for a service, check up and up date and says that all is well. I would be interested to hear if the problem is universal. I would also be grateful for photoshop correction advice, in easy stages if possible. I have had help on this subject a year or so ago but thought a revisit would help.
    In very clear weather the camera records obvious banding in the sky. I don't know if its because of clipping in the blues or an inability for the sensor to cope with extreme levels of blue. It records the same on jpegs and raws, so has little to do with compression, and shows on prints big time. Day time or evening shoots don't make much difference only the clarity and saturation levels of blue seem to be the main influence.
     
  2. Ian,
    Could you possibly post an example mentioning settings, lens, and eventually filter(s) used ?
     
  3. http://www.photo.net/photo/11893191
    camera set on... A priority...24-70 lens though it makes little difference..uv filter again it seems to make little difference.
    f22 at 1/15 iso200 though again I can get this result with other settings etc
    I hope you can find the photo from the link
     
  4. Color banding? Or those lines across that I can faintly see on my monitor?
     
  5. I cannot check it right on my monitor (it is a very bad office one), but I`d say it`s posterization. Higher color bit depth will help to avoid it. The color space used could also be the culprit of this issue, thought.
    Also, a low dynamic range pic could show this issue after processing. Maybe you should check the way/conditions you take your pics, and your processing way.
    I bet it`s not a D700 fault.
     
  6. "I bet it`s not a D700 fault."
    Too late for editing, I should have checked it before writing. I have already remember a straight pic of mine with an extreme posterization efect... but I don`t recall if it was taken with a D300 or D200.
     
  7. If you are shooting in just jpg, could it be the quality is set to low and the jpg compression set to max?
    looks like it's trying to minimize the number of colors and thus save space.
     
  8. Camera is set to highest possible quality jpegs and I also take raws at the same time
     
  9. Does look to be posterising on the image.
    Interestingly, the initial pic shows horizontal bands, but when you view at 100%, they disappear completely, but slight arc banding is visible (far less noticeable than the horizontal banding though - so not sure this is the problem you are referring too).
    It could be more to do with the display settings on the monitor as much as in camera processing.
    Do you have full exif data for the shot ? were any other processing options set in camera ? (ie vivid colour etc).
    Martin
     
  10. Ian, does it show in prints as well?
     
  11. Ian: Interesting. If you say you've seen in in RAW as well, it's probably not just the JPEG engine. Are you capturing in sRGB or AdobeRGB? When RAW processing, what colour space and bit depth are you using? If you're hitting the bounds of sRGB, you might have to reduce the saturation a little in the final image for final display - it might be worth trying this in RAW and seeing whether you still get banding. Or you could be right and you're hitting the sensor (or image processor) limits.

    If it makes you feel better, I have a Windows Mobile phone whose camera app crashes whenever there's too much blue in the image (AFAICT). Fortunately I live in the UK, and it's not an issue too often...
     
  12. If you could post the original raw file somewhere for people to download, that would make it an awful lot easier to diagnose.
     
  13. Its the arc banding that is giving me trouble. Camera is set to default more or less and set on neutral, rgb makes no difference. There is in my opinion nothing I can do in camera that will cure this, furthermore it is definitely originating in camera and shows on prints big time. Editing a 16bit raw does not cure it either. It needs some very clever editing skills to sweep it away and leave nothing but clean graduated blue skies
    As a follow on from my earlier post I have had to start a new one because for some reason I couldn't get it to accept my reply.
    So as a sum up I can say with 99% certainty that the bands of noise arcing across the sky originate in camera regardless of the setting rgb, raw or otherwise. It only occurs when and I'm guessing here, the atmospheric conditions give rise to high amounts of blue. I'm surprised no one else has experienced it as I have witnessed this phenomena on TV programmes showing sunsets where there is a strong gradation from the yellows to the blues of the upper atmosphere.
    What I don't know is if my Nikon camera is unique in failing in these conditions and how I cure it quickly and easily in photoshop.
    see link to photo
    http://www.photo.net/photo/11893191
     
  14. This may be a dumb comment, but are we 100% sure it's not an optical illusion of some sort ?
    On my monitor I cannot see the banding at all, and the arcing lines I can maybe see but not sure.
    Banding of colours I HAVE seen before, but on my TV, when I suspect the colour range is insufficient to graduate the colours accurately.
     
  15. Are you using D-lighting? If so, that could be the problem. I've noticed that using D-lighting causes very noticable posterisation in the camera's review histogram, so I refuse to use it. I shoot RAW all the time anyway, so I don't really care what happens to the JPEG version, and the tone curve can always be altered in ACR or PS at 16 bit depth.
    Try turning any D-lighting settings completely OFF and shoot RAW. I'd be very surprised if the RAW files showed any sign of posterisation.
    Another thing to try is just running the colour-picker over the apparent banding in PS. If it's real, then the colour picker will show up the step. Some LCD monitors only work at 6 bit quantisation, and so cannot show smooth gradations properly. I was heartbroken when my faithful old glass CRT monitor finally packed up and I was forced to buy a crappy LCD display.
     
  16. "I'm surprised no one else has experienced it... "
    I`m not an expert, but to the best of my knowledge it is normal on digital cameras. The deeper bit depth the lesser "step" effect. It not only happens to clear blue skies, but to e.g. flat dark studio backgrounds.
    I`ve been checking some of my D700s "beach" pics (the very few I have on this computer), and I hardly can detect it... although I don`t use to shoot landscapes or uniform flat areas. I use to avoid too much sky.
    "I don't know is if my Nikon camera is unique in failing in these conditions... "
    I bet it doesn`t. I don`t think so.
    "... how I cure it quickly and easily in photoshop."
    Just add noise. I tend to think that it should not happen at high ISO settings.
     
  17. http://www.photo.net/photo/11894832
    camera is set to Auto D lighting, and standard or sometimes neutral
    Banding arc is much less visible on a raw taken through photoshop
    and converted to a jpeg see link BUT if you look closely there is still banding
    and rest assured it will show on a print
     
  18. Oops - I've just looked at the full size image (there was separate posterisation in the small version).

    I think you're running into the limits of 16777216 colours: you have a slowly-changing gradient across a relatively small number of colours. 256 shades of each channel isn't as many as you'd think: the eye is very good at noticing a small discontinuity if there's a large area of one colour with another very similar colour next to it. If you shoot in RAW, use ACR outputting to a 16-bit (per channel, 48 total) intermediate, then convert to 8-bit (per channel, 24 total) with dithering enabled in colour settings, Photoshop should add some dithering to approximate the full (12- or 14-bit) range of the sensor data. That ought to help for printing, unless the printer tries to add some sharpening and accentuate the boundary (and if you can't print directly from 16 bits). When you convert it to a JPEG, you'll probably get the banding again, because JPEG isn't designed to handle high frequency changes with small colour differences - dithering works because the eye can't see it very well, and JPEG compresses by throwing away things the eye can't see. Short of saving a dithered PNG, which will be *big*, I don't have a good solution. Except possibly JPEG2000 and saving the 16-bit image, at least.

    As for fixing a JPEG once it's in this state... it ought to be possible to do something to detect an approximation to a gradient in 24-bit colour space, smooth it in 16-bit per channel, then do something with the result. It may be simpler than that - you could just convert to 16-bit, select the sky, apply an enormous blur filter to it and go from there.

    To quote Knuth (a famous computer scientist), I've only proven this correct - I've not tried it. But I hope that helps - it's at least worth a try.
     
  19. The only way to find out if the posterisation is present in the raw files is to analyse the data.
    The second best approach to the original data and easiest is to convert NEF to 16bit tif files and then to analyse the data by looking at the numbers. Or using some quantitative software like "imageJ" (free software) that will allow to show line profiles of the data along a selected line in the image.
    Until proven otherwise I bet that the banding is some posterisation either during post processing or when inspecting the images on a monitor . A "typical" monitor can only display 8bit resolution per color channel. One can easiliy run out of color resolution. And you are not the only one to see this - you are in a large number of people :)
    So even if your files are perfect using a monitor may downsample your perfect sky into visible steps of blue.
    The above suggestion to look at prints is a valid one - if the effect is not visible on prints it is the monitor or post processing for monitor viewing. But if the effect is visible on prints we still do not know and one has to quantitatively analyze the images.
     
  20. Let's address it as either a possible capture compression, raw processing, color space, output setting, or possibly a display issue.
    When you shoot your NEFs do you have the camera set to compressed, uncompressed or lossless compressed mode?
    Are you shooting 12 bit or 14 bit per channel NEFs?
    When you have imported the photos into your raw processing program (which one?) what are you doing to the data, what processing steps? When you are are ready to output as JPEGs or TIFFs, what color space (sRGB, Adobe RGB(1998) or ProPhoto RGB) and bit depth are you using to create the rasterized file in?
    Finally you may be seeing something on your display that I can barely see. Which display are you using, how is it calibrated and profiled, and what are the settings you are calibrating to?
     
  21. I hope this answers.....
    lossless compressed mode
    14 bit
    ACR
    Adobe RGB 16BIT
    Saved as a tiff 16 bit
    display....I have seen the image on 4 displays including iMacs all show it the same
    and on a A3 print
    thanks for all the help on this
     
  22. Try ProPhoto RGB instead of Adobe RGB(1998), ProPhoto has a much larger blue gamut.
     
  23. I use a D700, have for a couple of years. I've never had this problem in clear blue skies, of which I have many photographs. You have my sympathy, which isn't going to help a bit... t
     
  24. Ellis, the size of the blue gamut is not the issue. In fact, with a larger gamut, the visual difference between adjacent levels (especially at 8 bpc) becomes noticeably larger, making posterization worse, not better. Historically, this was exactly the reason cited by some early opponents of ProPhoto. Of course, at 16 bpc, this is almost a non-issue unless one is really stretching out some small differences in the sky to make it appear more visually interesting.
    Tom M
     
  25. So why not look at the data?
    1) Download and install ImageJ
    2) Convert a NEF file into 16bit tif in your "normal" way.
    3) Crop to an area where you see the effect the most.
    4) load the cropped image into ImageJ
    5) Place a line at right angle across the "steps"
    6) Create a profile plot along the line
    7) Look at the profile if it shows any visible steps.
    This will give you a starting point to either go upstream or downstream in image processing to identify the cause of the problem.
     
  26. Ian, I mentioned above a shot I did time ago with a coarse banding issue... well I have found the pic. It was taken in summer 2008 with a D700, and originally downloaded&cheked in another computer. I`m currently at home on a 8bit iMac screen and that bizarre posterization is now not visible.
    Just a question; does this "banding" you`re experiencing have the very same look in all your "banded" pics?
     
  27. @ Tom Mann, but you don't use ProPhoto RGB at 8 bpc.
     
  28. @Ellis - You suggested he try ProPhoto, essentially suggesting that it can improve posterization. In fact, it can not do this. It can only make posterization worse ... a lot at 8bpc, less at 16 bpc. This is why it is hardly ever used at 8 bpc.
    Tom M
     
  29. CMOS sensors tend to do that. Nikon has done a good job with their sensors, but sometimes, in weird conditions, banding (actually, posterization) happens. I wouldn't call it normal, because it's hideous, but on images which are basically just monochromatic blue... It does happen. Not everyone sees it, there are differences among cameras, atmospheric conditions etc.
    But D700's are usually (way) better than that. Try to communicate with Nikon again. Maybe you'll have more luck next time. They definitely should fix it.
    In the meantime, I believe RawTherapee could help achieve better results. Also Nikon View/Capture NX. ACR is the worst for posterization, IMO.
     
  30. I'm going to bypass the obvious technical conversation above.
    There IS an easy fix for mild posterization. Add luminosity noise to the banded areas. Make the radius bigger until the posterization goes away. Noise is not always the enemy. It visually smooths out gradients and adds perceived sharpness.
     
  31. @Jose-banding looks similar in each photo in that it follows an arc in the steps of gradations in the sky and only occurs when using a wide lens.
    Its at its worse when shooting a sea scape with the sun just below the horizon and a clear blue sky above.
     
  32. That`s the reason I was looking for that "beach" pics... I suspected it had to be more noticeable under that conditions.
    Better safe than sorry... maybe unwittingly, I use to follow this rule. For several reasons I rarely include more than a very thin sky line in the top of my framings when shooting wides. I also take extremely accurate metering when shooting flat areas.
    Again, I`m not an expert, and there are things struggling my head:
    • Would a higher pixel resolution solve this problem? Maybe higher resolution could break that awkward aggregation.
    • Higher bit depth should solve the problem, though. Maybe 12-14bit is at the edge under this precise conditions.
    Anyway, on my own pics I hardly can detect this banding on some pics. Under a very deep inspection in most pics I almost have to guess that the issue is there... in a good monitor. Bad monitors are treacherous. Maybe my camera settings help:
    • I always have all the camera tricks off (D-Light., NR, auto-ISO)
    • I always shoot in neutral, (and I try to remember that sharpness is even lowered to neutral -1)
    • Always RAW - I convert it in NX2
    • Big prints always sent in TIFF format.
    Looking for tutorials, I have found this one (this guy also follow the better safe than sorry idea... ), and of course, a must-read cambridgeincolour tutorial, but I suspect you have already read them.
     
  33. @Jose... Thanks for the tips and links, I had not read those articles. Its all a little clearer now. I'll change some of the settings on the camera and play around with the exposure and see how I get on.
     
  34. I downloaded your image:
    http://www.photo.net/photo/11893191see link to photo
    and opened it up in Photoshop. I see no banding.
    I found if I view your image on PNET at normal size, it shows banding on my monitor but if I click on the image to enlarge it, there is no banding visible. The banding you are seeing may be limitations of your monitor.
    This image below is a deep crop and and inset of a very extreme crop - there is no banding.
    00XcKB-297907584.jpg
     
  35. Yeah but he has stated twice or three times that it is MOST noticeable on prints; so it isn't his monitor. Although calibrating the monitor might help the prints. (Try Spyder3Pro.)
    Definitely take off the D-Lighting, which was mentioned once but seemed to get lost. And Nathan B's point on noise is very apt. See if your NR settings are too high.
     
  36. As correctly pointed out earlier in this thread, posterization in excess of the amount present in *any* original file (for example, a gradient known to be as smooth as the quantization steps in the color space in use permits) can be caused by various problems such as monitor inadequacies, wide gamut color spaces, transformations back and forth between different color spaces or different bit depths, etc. As Elliott correctly pointed out, these problems can be exaggerated at certain magnification ratios.
    However, it is quite clear to me that posterization / banding is unmistakably present in the file under discussion. It's not huge, but it's definitely there. An easy way to distinguish posterization in the file from the other sources of posterization that I mention in the previous paragraph is to enhance the contrast of the original image data by such a large amount that any posterization seen in the result is much larger than any possible amount of posterization induced by monitor inadequacies or the other problems mentioned above.
    Because the image under consideration is a gradient, simply increasing the contrast (say, with a contrast adjustment layer) will make the lower RH corner of this image go completely white, while the upper LH corner will go black, limiting the region in which you can detect posterization to a small diagonal band in the middle of the image. A better way to bring out posterization problems in such gradient images is to use a tool which enhances local contrast. There are many such tools / approaches that can be used to enhance local contrast. I happened to use the "Tonal Contrast" tool in the Nik Color Efx pro plugin.
    The result is shown in the attached image.
    One can confirm that the pattern seen after the application of strong local contrast really is coming from the original image and is not an artifact of the enhancement process is by intentionally distorting the original image. For example, one can use the liquify tool to distort any banding present in the original and then apply the same enhancement process to the distorted image. I have done this, and the bands in the contrast enhanced image are distorted exactly as one would expect, thereby eliminating the possibility that the bands seen after the application of local contrast are caused by some unknown artifact of the enhancement tool used.
    That being said, this discussion would be enhanced greatly if, as already requested, the OP would post the RAW file for this image (or email it to some of us, etc.). The real issue that seems to be troubling the poster is whether there is something wrong with his camera. There is no way to tell this from the file he provided because banding of blue skies in 8 bpc JPGs is as common as dirt and says absolutely nothing about his camera, sensor, etc. Having access to the corresponding RAW will settle this one way or the other.
    HTH,
    Tom M
    PS - @Elliott - Cropping and changes in magnification, such as you have done, are not the proper way to detect posterization in the original image data. These changes will only change the amount of additional posterization caused by monitor inadequacies, problems with the algorithm PS uses to display images at different magnifications, etc.
    PPS - @Ian - You have a bit of bit of dirt or a water spot on your sensor (ie, in the lower RH corner). This local contrast technique brings out all sorts of image problems that you may prefer not to know about. ;-)
    00XceT-298187584.jpg
     
  37. Tom, I was addressing the banding issue only, not the posterization. I had a camera with banding and in my opinion, there is no banding in this image,
     
  38. Hi Elliot - Posterization is, by far, the most common cause of banding. However, one needs a nice uniform area of solid color (e.g., a sky) to see banding, even if strong posterization is present. Because the image under discussion in this thread is a typical blue sky gradient, I made no attempt to distinguish between the two concepts in my post.
    To clarify my previous post, I see banding due to posterization in this image. I can see this on my monitor before contrast enhancement at (almost) any magnification. It's not a large effect, but it's definitely present, and it becomes trivial to see once the contrast is enhanced.
    Would you care to elaborate on the distinction you are making?
    Cheers,
    Tom M
     
  39. I will post the RAW files. What is the best way to do this?
     
  40. Earlier today, Ian sent me two RAW (ie, NEF) files from his d700.
    The 1st (#5526) was a long (13 second) time exposure with a stretch of placid ocean in the bottom half of the photo and a gradient sky with the remnants of a sunset in the upper half of the image. In ACR, with all sliders set to the Adobe CS5 default values, the histogram showed an extremely compressed range with abrupt drop-offs of each of the primary colors. Since this image was taken under unusual conditions (ie, 13 seconds), wasn't the one under discussion in this thread, and had an extremely unusual histogram for a RAW file, I'm not going to discuss it extensively at the moment.
    However, I will comment that since this is essentially a gradient image with little horizontal detail, spatial bands must be produced at each of the sharp edges in the histogram, and the range of tonal values must be greatly expanded to convert this data to a visually pleasing image, it's obvious this file is ripe for posterization, so I'm not going to discuss it further. The unusual histogram is appended below.
    Tom M
    00XdCa-298823584.jpg
     
  41. The second file NEF file (#5347) that Ian sent me was the source of the JPG image under discussion in this thread. It appears that this file was, at minimum, cropped to produce the JPG image under discussion. The shooting conditions (1/15 second vs 13 seconds) were more conventional than the 1st image. Its histogram was spread out a bit more than image #5526, but, because it contains only sky, was somewhat restricted to a relatively small range of tonal values in each of the primaries. Like the other image, there are some relatively sharp drop-offs in the separate histograms of each of the primaries, but not quite as sharp as the other image. The combination of somewhat restricted range and fairly sharp drop-offs will tend to produce posterization, but not as bad as other image. The histogram of this image (as opened) is shown below.
    00XdCq-298829584.jpg
     
  42. Next, I had ACR convert the image to a 16 bpc file in the sRGB color space for further investigation using CS5. By eye, I could see much less banding in this version compared to the JPG version posted at the start of this thread. To enhance any banding that might be present, using the exact same settings as I used to process the JPG image, I used Nik Color Efx Pro's Tonal Contras filter to enhance local contrast.
    The result is attached below. There may be some suggestion of weak banding, but it is clearly much less than the corresponding processed image for his JPG version.
    My conclusion is that there is nothing wrong with his d700, but rather, something happened in his conversion of the original file to encourage some relatively minor banding. This could have been something as simple as expanding the tonal range after conversion to an 8 bpc file instead of at full bit depth in ACR. In any case, from the looks of the NEF's he sent, I don't see anything obviously wrong with his camera.
    More later. Gotta run.
    Tom M.
    00XdCz-298839584.jpg
     
  43. I've started to get more critical of Photoshop and ACR lately. At one time I believed the hype that PS was the DADDY of image editors, however I've had one or two nasty surprises that make me look a bit sideways at it now.
    Surprise number one was finding severe posterisation in the shadows of all my images when certain combinations of colour space and monitor profile were used. This seemed to be real, inasmuch as PS was actually damaging the image data when opened and confirmed by the colour-picker readouts. Meanwhile "crappy" PaintShop Pro didn't alter the image data at all unless I told it to. The only workaround I could find was to basically turn Colour Management in PS completely OFF.
    Surprise number two came just the other day. Opening a 14bit NEF RAW file in ACR, I decided to adjust the saturation of greens slightly (by about +10). When the image was imported into PS at 16 bits I noticed there was some awful posterisation between the yellow and green shades on some backlit leaves. I tried the same adjustment using PS itself, rather than ACR, with a subtley different but still posterised result. Adjusting the crossover between colours in Hue/Saturation didn't help. More surprisingly, I got EXACTLY the same result by doing the adjustment at 8 bits/channel instead of 16. Makes me wonder if 16 bit support in PS is just bullsh*t.
    Roll on a 16 bit version of GIMP.
     

Share This Page