how to icc profile a camera?

Discussion in 'Digital Darkroom' started by nino_loss, Dec 20, 2010.

  1. What I tried till now resulted in profiles that had at best an peak DeltaE of 17-18 and an average of 7-8. Most of the times the profiles are to saturated and overly reddish... I use Argyll now, after worse results with basICColor Input.
    Here is the Capture One tethered studio procedure and what I tried:
    - evenly illuminated ColorCheckerc24 (double checked evenness)
    - tried different exposures with and for different applied curves
    - WB with BabelColor target
    - applied "color view profile (former "Phase One - No color correction)" and either standard/linear
    - no curves / applied curves to match CC24 gray patches "ideal" values
    - everything else zeroed, including sharpening and NR
    - processed 16bit TIF with "Embed Camera Profile"
    - in Argyll I use
    • scanin -v NameOfMyFile.tif ColorChecker.cht ColorChecker.txt
    and than
    in basICColor I also used the... basic settings ;-) keep camera gray and no absolute white
    thanks in advance for any help with this
    regards
    nino
    P.S.: I basically want to see if I can get decent results with a custom ICC profile, before investing even in a ColorChecker SG. I couldn't get hold of a trial for the ProfilMaker or the MonacoProfiler, not even of the InCamera software.
     
  2. I think Xrite has a software called Passport or something. It just needs the std cheaper Colorchecker 24 patch (?) but it might support the more $$ ones too. The software's free.
    Thou I think, for a digital camera profiling you need to do it in each setting, not much good for out and about photog but v good for controlled setting like weddings, where you might have one inside and one outside when light might not change a lot. So you can apply the profile thru out all the shots.
    FWIW I have the Incamera Plugin, not sure if I have a older version, I used the 24 patch Colorchecker on my flatbed scanner - the results are horrible.
     
  3. Thank you Ray, but x-rite's CC passport does not produce ICC profiles.
     
  4. If you are going to create an icc prcompliant profile for a specific camera body, what you reall yare doign is creatign a profile for not just that specific indivual body, but also the lens you used t o photogrpah the target , and the specific lighting condition (time of day and atmospheric condition if usign sunlight; light source(s) + Modifier(s) if using artificial light.
    The profile will be inaccurate as soon as you change the lighting and the lens.
    I have the SG ColorChecker Target and have profiled cameras with it in the past . I would rather not do this again. You might want to see if Thomas Fors script for ACR is available but I doubt it works on any version fo ACR past maybe ACR 2 or 3.
    A better approach is to use the Dng profiling software. there is the aforementioned ColorChecker Passport software from X-rite but you should also check out the more powerful and customizable Adobe DNG Profile Editor: http://labs.adobe.com/wiki/index.php/DNG_Profiles:Editor authored by Eric Chan.
    With either of the DNG profile editors it is best to make a dual illuminant (daylight/flash & undimmed 3200˚K Quartz-Halogen) profile.
     
  5. Ellis,
    I have and use the CC passport (and before it was released I also used the script). I am aware of what it can and can not do.
    I need icc profiles for accurate color reproduction in a controlled studio lighting setup. I know that I will do this for every camera and every lens etc...
    If you have done this before, maybe you can help me or point me to documentation. Please do see my initial post on what I have already done and which software I used.
     
  6. For accurate color reproduction as in product and fine art shots, you're better off creating a DNG profile as Ellis pointed out.
    But I wouldn't create a dual illuminant version for your purposes, just the 6500K table. From my experience with the dual illuminant profile, it generally adds a bit more blue to bright red as well as adding an overall red tinge to warm pastels depending on which way your default color temp leans toward green or red. It's very subtle.
    From my experience getting exact color rendering of objects right next to my computer that I've shot in Raw, applying the custom DNG profile, editing by eye and saving the setting and applying to the rest in a studio shot series is MUCH quicker and less hassle over going the expensive software route. I think DXO Lab's site has an ICC profile creation package but it ain't cheap.
    The DNG profile as I'm assuming whatever competing ICC software out there does is add small tweaks to shadow definition and correct for all primary color errors.
    Whenever I apply my DNG profile to outdoor shots or shots I didn't create a profile for different lenses and lighting, I find I can fix quickly using ACR's HSL panel while getting the right look to the color temp.
    Shooting in a light controlled studio should be a breeze getting accurate color reproduction.
     
  7. Just to add another issue you're going to run into shooting a variety of objects or artwork is that each subject being shot in the studio may or may not bring their own oddball spectral reflectance characteristics not calculated for within in either DNG or ICC profile. For instance a dye based purple fabrics may show up blue while another synthetic purple fabric will look OK. You can always correct it using the HSL panel. It's a lot easier.
    Don't know what you're shooting so your mileage may vary.
     
  8. Tim,
    Thank you for your comments.
    I have to do some oil paintings and composite art work.
    I tried the DNG rout with the DNG Profile editor. The latter is still a Beta product and does not work properly on my win7x64/16ram/ATI HD5770 studio workstation. Even if it would work, I see a few practical GUI problems (In the "color tables" tab, I can not work on a spot by the numbers, 'cause once the picker is not over the image anymore the numbers disappear. To find the same spot again is wishful thinking). Once those issues resolved, I am quite sure it'll be the way to go.
    For the moment x-rite sells two professional software packages for profiling. They are icc only and about USD 2.000.
    Shooting in a light controlled studio should be a breeze getting accurate color reproduction.​
    I wish you where right! Here is a fine little introduction to the problem: http://www.betterlight.com/downloads/conference06_notes/color_Accuracy-ppt.pdf
     
  9. The ColorChecker DC will do a better job than the 24 patch target which wasn’t designed to build ICC camera profiles. But ICC camera profiles are fraught with frustrations. They are output referred, meaning they describe how a raw file has been rendered which is putting the cart in another city, not before the horse. If you have an Adobe product for raw conversion, you might try DNG profiles (Passport from X-Rite is free).
     
  10. Thank you Andrew for keeping replying me. You continue telling me to change to Adobe. Well, as I tend to listen to those who should know better, I took an hour today to look into my DNG possibilities. Here in the studio that would be Lightroom/Bridge/Camera Raw/PS. I really dislike Lightroom for various reasons, but I could reduce it to thethered shooting only and than work with Bridge/Camera Raw/PS. I really like Bridge.
    Two things though:
    1) The DNG Profile Editor is still a beta and it is not stable here (win7x64). (BTW in the "Color Tables" Tab, when I want to edit colors by the numbers, once I move the cursor/picker away from the spot that I want to measure/modify the numbers change/disappear... So how to edit by the numbers? I don't want no visual adjustment.) So how would one build a top quality DNG profile?
    2) How come that a thing like Phase One-Mamyia-Leaf ignores DNG and that they continue to say that any real pro uses ICC? X-rite still sells for just under USD 2.000 their ICC legacy products ProfileMaker and Profiler saying
    If you’re a professional fine art or commercial studio photographer with a discerning eye for color, the highest quality profiles are a must. When paired with an i1 measuring device, ProfileMaker™ 5 Platinum helps you connect each step in your color imaging workflow by creating and editing high quality, reliable ICC profiles. Achieve the most accurate digital color with highest confidence that it will remain consistent on each input, display and output device.
    (source http://www.xritephoto.com/ph_product_overview.aspx?ID=1211&catid=111)​
    So that's not true anymore? They forgot to update their new website?
     
  11. So that's not true anymore? They forgot to update their new website?​
    That’s marketing speak. Its not necessarily more or less true today than back when it was written. In a very controlled situation like copy work setup, an ICC camera profile can indeed work very well because you are profiling a very specific and consistent behavior, much like profiling a scanner. If the illuminant changes, the dynamic range, the exposure and so forth, all bets are off.
     
  12. Thank you Andrew! So for a real copy work that would still be better. I see you ignore my question about Phase One et al insisting on ICC.
    So we wait for the new x-rite software packages in order to be able to get good profiles?
     
  13. So we wait for the new x-rite software packages in order to be able to get good profiles?​
    Not ICC camera profiles, it doesn’t make em.
     
  14. That's what I meant Andrew: We'll wait for the new x-rite software to get good profiles, which are... DNG
    I hope Phase One will somehow jump on that train!
     
  15. There is nothing new in the new software for profiles (ICC or DNG). DNG profiles are built using the current Passport product.
     
  16. and the DNG Profile editor? That's what might save the day, because as is, the DNG passport profiles are not... good. I today had to take an image of face mask painted in skin like tones with big pieces in dark orange and red on it. Although I build the profile from the beginning with the same attention for minute details, still the orange was way oversaturated and together with the red and the skin, off in terms hue. Noticeably so. Now I could tweak around, but tweaking here, untweaks there. To make things simple I also had real human skin on there.
     
  17. The DNG editor is an Adobe product. You can edit Passport generated DNG profiles using it.
     
  18. so let's hope the DNG editor will come out of beta soon... and do wonders.
     
  19. Although I build the profile from the beginning with the same attention for minute details, still the orange was way oversaturated and together with the red and the skin, off in terms hue. Noticeably so. Now I could tweak around, but tweaking here, untweaks there. To make things simple I also had real human skin on there.​
    What's your light source? And can you post this image showing before and after profile application.
    What you just indicated sounds very similar to my experience shooting a wide range of objects including real skin tone under two 5500K 100watt Ottlight CFL's using a custom DNG profile for those lights. The profile doesn't fix all colors for every different object I shoot under these lights. It requires tweaks to specific colors mainly reds and oranges.
    What I find is the appearance of contrast perceived by our eyes viewing actual objects lit under various lights will not exactly match the DNG profile's which have color tables established by a fixed exposure induced contrast ratio. Its contrast is based on referencing a flat target whose numbers are defined by measuring with a spectro at the factory and calculated within a linear space. It gets you in the ballpark establishing individual colors without noticeably changing contrast and luminance.
    Different objects will give off there own specific color characteristic according to the light source and contrast ratio. It's the same problem I had profiling a scanner using an IT8 silver halide target. Once I built the ICC profile and applied it to some artwork or fabric design I scanned, the profile made the scan of those natural objects look like crap while all silver halide prints matched exactly.
    The digital camera is a lot more forgiving capturing natural objects (not silver halide prints) because it's not strapped to using a scanner's bluish fluorescent light source combined with the limited and/or specific spectral reflectance characteristics of a silver halide IT8 target. You're shooting a sculpture who's pigments reflect back under a fixed light source a specific luminance, hue and saturation level not calculated for in the X-rite's target.
    I can usually fix those kind of color errors you describe by reducing/changing contrast induced tonal distribution and tweaking HSL and color temp. If you're going to shoot a wide range of different objects made from material not included in the X-rite color chart, you're going to have to do some tweaking.
    Just a tip but I'ld suggest not making the image look dramatic with beefed up contrast. As an example ever notice when you increase contrast on an image in Photoshop or ACR/LR, you get luminance, hue and saturation shifts? Sometimes applying contrast with a wide sloping curve tweak can fix these kind of contrast induced color errors instead of just using the contrast slider.
     
  20. Ellis,
    my major concern is the 5D2's reds (orange, pink...), but for copy work I also need the HDH40 to be spot on. For big things I rent PhaseOne, or Leaf as a matter of fact.
    Tim,
    for my studio work, I use natural day light (which is abundant and extremely predictable where I am) and my Profoto D4 with Profoto and Elinchrome light-shapers.
    What you describe sounds all too familiar, especially contrast tweaks giving HLS shifts. I only disagree with
    I can usually fix those kind of color errors you describe by reducing/changing contrast induced tonal distribution and tweaking HSL and color temp. If you're going to shoot a wide range of different objects made from material not included in the X-rite color chart, you're going to have to do some tweaking.​
    I think, one can tweak very well to obtain PLEASING results. Only, I very often need more ACCURATE rendition. How do you get accurate with the DNG Profile Editor? How to edit by the numbers?
    An example, maybe you can help me:
    In this example (not real, just a quick screen shot), I photographed a ColorChecker, picked on a few colors. Now I can tweak them, but ONLY VISUALLY. There are no numbers showing anymore. In the Advanced Color Editor of Phase One's Capture One RAW converter, there are the tools for that, there are "color read out" points which you can set on the picture. They change as you tweak. Why can't there be such a thing here?
    00Xth1-313681584.jpg
     
  21. So here is a video that explains the, to me, more convincing and practical approach, which is ICC profiling:
    http://www.luminous-landscape.com/reviews/accessories/coloreyes-commercial.shtml
     
  22. BTW you could than use the ICC profile in Capture One and still tweak and edit the profile (if need would be). But th in C1, as mentioned you have the ability to set "color read out" point, so you actually see what you are doing.
    So here is a video that explains the, to me, more convincing and practical approach, which is ICC profiling:
    http://www.luminous-landscape.com/reviews/accessories/coloreyes-commercial.shtml
    P.S.: Don't miss that video!
     
  23. None of this work, profiling with DNG profiles or otherwise is “accurate” (meaning colorimetrically correct values). Its all about output referred, pleasing (expected, desired) color. see:
    http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf
     
  24. It's understood that this paper from 2005 and it's arguments you mention, where known to the participants at Sheridan college in 2007 in this very informative video on that page. Michael Reichmann insists in this video about this color eyes icc camera profiling technique that it convinces him.
    Also, "accurate" is meant in terms of final output, i.e. a critical client compares the repro to the original and agrees that it just looks like the original.
    So here is a video that explains the, to me, more convincing and practical approach, which is ICC profiling:
    http://www.luminous-landscape.com/reviews/accessories/coloreyes-commercial.shtml
     
  25. In future there will maybe be other possibilities, like spectral measurements of the scene integrated in a way that the output matches when measured with a spectrophotometer. If x-rite-Adobe manage to do this that'll be an other story, obviously. Maybe that's the thing thats coming now?
     
  26. Also, "accurate" is meant in terms of final output, i.e. a critical client compares the repro to the original and agrees that it just looks the original.​
    Meant by whom and accepted by whom? The above cannot be defined as accurate as there is no measure of success or failure and the degree of each. Its totally subjective. That is exactly what colorimetry is designed to do. What you prefer to call accurate is pleasing or visually matching color. The values that produce this in no way match the values predicted by colorimetry, a science that is designed to provide an exacting and verifiable means of specifying something accurate.
    So here is a video that explains the, to me, more convincing and practical approach, which is ICC profiling:​
    The video or the process as tested by you have convinced you of the practical approach?
     
  27. In future there will maybe be other possibilities, like spectral measurements of the scene integrated in a way that the output matches when measured with a spectrophotometer. If x-rite-Adobe manage to do this that'll be an other story, obviously. Maybe that's the thing thats coming now?​
    This is exactly the right way to do this and has been proposed by a few color scientists (notably Eric Walowit) on the ICC Digital Photography Working Group of which I’m a member. The idea would be to measure the illuminant of each capture and along with the known spectral sensitivity of the chip, build an on-the-fly “profile” (probably as EXIF data). Its possible today but we need quite inexpensive and small Spectrophotometers built into the camera itself which I suspect is the biggest hurdle in implementation.
     
  28. so that's a quite far removed future, non?
     
  29. Meant by whom and accepted by whom? The above cannot be defined as accurate as there is no measure of success or failure and the degree of each. Its totally subjective. That is exactly what colorimetry is designed to do. What you prefer to call accurate is pleasing or visually matching color. The values that produce this in no way match the values predicted by colorimetry, a science that is designed to provide an exacting and verifiable means of specifying something accurate.​
    I hear, but let's say, I need to make the copy of an oil painting, in the end I will hang it in my standardized contractual ISO certified (and I don't know what else) viewing booth, next to the original, and upon close inspection I and later the client, will agree that this matches. So they look alike, but they don't read out the same values?
     
  30. So they look alike, but they don't read out the same values?​
    Correct. You can measure the lab values of the original and the lab values of the copy. They could match and the Lab values can be greatly off. The colorants are totally different. The paper is totally different. That they match visually is called a metameric match (often incorrectly called metamerism and used in a negative fashion). (2) samples with different spectra compared to each other with a given set of viewing conditions, produce a match. Metamerism only applies to two color patches when they are compared. It is not correct to refer to one color from a given ink or paper and say that it suffers from metamerism. A "meterameric pair of color patches" means that they appear to match under a given illuminate. However, they may not appear to match under another illuminate.
     
  31. nino, I don't seem to have as much trouble color matching objects lit under Ottlight CFL's as you are using a mix of actual daylight (from a window?) and artificial light which I suggest you stick with artificial light in a studio setup. Mixing two types of light can throw off sensors. And besides, daylight changes throughout the day offering up even more inconsistencies. You want a single stable type of light to build the profile from. This is a basic copy stand procedures even shooting with film.
    Remember digital or film camera technology was never designed to be a copying machine. It's designed to capture wide gamut and dynamic range scenes and compress them into the gamut and dynamics of a print or transmissive display. As Andrew pointed out it would be great to have the luminance, hue and saturation readouts at the sensor level because some scenes can overwhelm the electronics of a sensor in unpredictable ways to where luminance levels don't correspond to saturation and hue levels which seems to be the case concerning your capture of the colors in the face mask.
    Sunlit orange pomegranate flowers always overwhelms my camera's sensor regardless of exposure compensation. I can fix it shooting Raw because the digital sensor is best at recording luminance, not so good with saturation. It's similar to what happens viewing stain glass windows in a church where some color's luminance and saturation stick out like reds and oranges compared to certain blues, greens and browns. The pigment mixes used to make the stain glass contribute to this effect as well. Same thing happens to Bayer RGB filtering and can be different depending on camera brand and model.
    Another thing is don't use DNG Profile Editor's selective color editor shown in your screenshot for correcting color patches on an X-rite color target unless you don't mind a lot of noise in hues close to the colors your are editing. It happened to me. That editor is meant to be used on individual images like your face mask capture.
    It's also meant to correct for a series of images that have similar colors like the face mask image. If this is the only object that induces these types of color errors, it's best to create a modified profile of the original you started out with and save it separately to apply only to those images having those colors similar to the face mask's. Otherwise it might induce color errors and banding/noise to other captures not like the face mask image.
     
  32. [...] A "meterameric pair of color patches" means that they appear to match under a given illuminate. However, they may not appear to match under another illuminate.​
    so than there is only visual match for all practical purposes? And how does the above mentioned idea make sens? The idea was spectral measurements of the scene integrated in a way that the output matches when measured with a spectrophotometer. You noted that
    The idea would be to measure the illuminant of each capture and along with the known spectral sensitivity of the chip, build an on-the-fly “profile”​
    Why should we need this, if anyway they only appear to match, and even that only as meterameric pair?
     
  33. so than there is only visual match for all practical purposes?​
    No, just the opposite. There can be lots of meterameric matches. Without this, we’d never see two different printers, or a printer and a display produce a visual match.
    Why should we need this, if anyway they only appear to match, and even that only as meterameric pair?​
    It would not solve all color matching issues but would go a long way towards describing how the capture device “saw” the scene and allow us to render it in a far more colorimetrically savvy way. Right now, you shoot a target under some fixed illuminant and build a profile. That profile has no idea you just left the scene and are using a totally different illuminant for one.
    This might sum the issue up best:
    Building a perfect ICC profile for a digital camera viewing real
    world scenes is only possible if the spectral response curves of the
    camera happen to be an exact linear combination of the human eye's
    cone response curves. Since there are no such cameras, perfect
    digital camera profiles are impossible.
    There are always going to be some objects that the camera "sees" to
    be the same color, and the eye sees as different colors. A profile
    cannot undo this data loss.
    Thomas Knoll.​
     
  34. Tim, thank you for your observations.
    In studio, I normally don't mix daylight with strobes, especially not when doing art reproductions. But, for example with food, natural daylight does wonders. The light flows in trough large scrimed windows, glass ceilings fitted with shades to direct the light.
    Regarding the example with the mask, I don't see what I can do, because tweaking does not give any usable results. We talk about red, orange, flesh like pink of the mask and real skin tones of a face, all in one picture. Any tweak I make ruins the human face... ICC or DNG color editors are useless for that. Only layers help here.
    But what if we talk about simple, like a typical red canon cast. Here also the problem in a simple portrait will be how to edit the colors. It does not help to take out the red, because it'll go missing in the lips and the skin. Tried a thousand times.
    Another thing is don't use DNG Profile Editor's selective color editor shown in your screenshot for correcting color patches on an X-rite color target unless you don't mind a lot of noise in hues close to the colors your are editing. It happened to me. That editor is meant to be used on individual images like your face mask capture.​
    I have noticed that too. The whole editing and tweaking can only be very very subtle, because the image quality suffers, especially if you have to print large.
     
  35. You're using strobes?
    I don't see how you can get a visual match on your computer using lights your eyes don't even see lighting the subject. A strobe is pretty quick.
    Have you tried using continuous artificial lighting?
     
  36. Yes, I tried shooting even large objets in my JUST viewing booth.
     
  37. Ellis,
    are you still there?
    Why did you ask me which camera?
    I answered you 5d2, HDH40 and rental leaf/phase...
     
  38. EDIT: deleted accidental double post
     

Share This Page