This was all SUPER enlightening, everyone! Thanks for so much engagement! I learned a ton.
2 Likes
Hi all,
Thanks for these detailed discussions. I had been successfully scanning B&W and color negatives and now slides, but the slides prove problematic. The slides are new (Fuji Velvia 100), developed professionally (not by a “bulk” lab) and look fantastic by themselves, but when DSLR scanned, the colors are way off and look crummy. In the pic below, the left is as scanned with a custom white point taken on a blank section of the film strip, and the right is after inversion in LR and using NLP to “develop” the “negative”, using LINEAR, Fuji WB, and no LUT. The right is (much) closer to what I see on the slide.
After reading the above and playing around, it seemed indeed the profile that is the culprit, so I decided to make a custom white balance and camera profile using a Colorchecker Passport and the X-rite LR plug-in:
Inside the enlarger head that I use for my scanning is a Solux halogen daylight bulb, a collimator, and a glass diffuser and I used that entire optical string to light the Colorchecker passport and take a picture with the same lens I use for scanning (Nikon AF 105 2.8D macro).
The thought process was that using this custom profile, generated with the same light and lens as used during scanning, I should see on my monitor exactly the same as I see on the slide with the same light source, and the specific color rendering of Fuji Velvia should be visible on my monitor…but…what I get is nearly the same as in the first picture above (which uses the generic Adobe Color profile for my camera), i.e. crummy with incorrect colors.
So now I’m confused. When applying the (linear) NLP 2.3 profile, the colors seem better, but the picture looks rather “flat”, and didn’t I just go through calibrating the entire chain, so shouldn’t my custom profile with white balance and camera calibration taken off the Colorchecker Passport using the light that also lights my slide during scanning not give me exactly the same colors as I see on my slide?
Any suggestions as to what I am doing wrong here? Your clearly expert inputs are much appreciated.
Arno
Arno, as far as I know - Nate can confirm one way or another - NLP is not made for processing positive transparencies. You should have a decent device profile, “scan” the slide, import it into Lightroom or Photoshop and process it in Adobe ProPhoto or similar wide-gamut colour space,.
Hi Mark, yes, I’m aware that in theory I wouldn’t need NLP for slides. I simply posted my question here because of the discussions above. In theory I should now have a good profile (apart from using a calibration target with many more patches) and my colors should be calibrated and correct, but they are not, so I’m puzzled and wondering what I am doing wrong. As it stands, my own calibration profile seems to be close to adobe’s generic profile, but both don’t give correct colors.
I read somewhere in these forums that one should use a “linear” profile when scanning slides, such as an NLP profile (if I understood correctly), and indeed the colors seem better with that, but it generates a rather dull and flat picture, and does not have the custom calibration that ideally should be possible. I simply used NLP on an inverted file since I noticed that it gave me a better picture, but I shouldn’t have to do that on a calibrated system, no?
Still puzzled…
Leaving NLP to one side, and taking your RH image as being close to the original, or at least much closer than the LH unmanipulated version, I must say that I’m pretty staggered as to how far off that LH one is. No reflection on your method, I’m just struggling to reach for an explanation. At various times I’ve experimented with different light sources for slide copying - flash, tungsten enlarger head, specialist daylight balanced fluorescent tubes, high CRI LEDs & an Ipad. The daylight tubes are in a Hancocks lightbox, no longer obtainable of course. I have two of these and I used your method of lighting the Colorchecker with them (one each side and in the dark of course) and using the Lightroom plugin to create a profile. As I explained in the thread I also made use of a custom profile made by a friend using an IT8 target.
The only reason I’m going into this is to say that in all cases the differences, such as they are, are subtle, hard to choose between them, or between each of these and the flash which you can probably take as a reference. There are differences, which is why I am interested in colour profiling as described by Mark in his article, but they are certainly not in the same league as your pairing here. So as I say, I’m perplexed as copying slides in my experience is far more straightforward than dealing with colour negatives which is where NLP excels.
I’m wondering, once you’ve created your profile in Lightroom using your Solux bulb and enlarger head does the colorchecker chart look right on screen? Compared to say with how it might look if you’ve done a profile using daylight or some other common light source?
Yes, I was also quite shocked how bad the unprocessed scan looked on my monitor. I shot some Velvia just to see how far I could push my digitizing setup, thinking slides would be straightforward. The difference in color made me think of what happens if one prints with the wrong color profile. I had a similar issue once when I sent jpg’s to a printing service and had accidentally assigned an Adobe RGB profile to the files, whereas the printing service expected them to be in sRGB.
After checking the slide scans, I noticed that my camera was set to Adobe RGB, but since I import RAW files, whatever color space is set in the camera shouldn’t matter. Maybe I’m completely stupid and wrong and I need to tell LR somewhere to expect Adobe RGB for the RAW files? I will try and scan some again with the in-camera color space set to sRGB but I’m unsure why that should matter with RAW files.
I hope someone else can explain, it won’t be the difference between Adobe RGB & sRGB though, at least not in my opinion. The main visual consequence of someone (a printing service say) interpreting that incorrectly is a difference in contrast and saturation rather than a profound difference in colour balance. You can see that yourself by ‘Assigning’ an image with the wrong profile in Photoshop, so assign Adobe RGB to an sRGB image, and vice verse. As you say, with RAW files it makes no difference at all.
You need a good quality transparent device profiling target which is back-lit with the same lighting that you use for scanning your slides. Once you scan the target, you then need to create the profile from that scan using a good input profile-making application. Once you have a good quality profile for the scanning device, the slides should scan with decent colour. Linear scans will normally be flatter and darker than you expect for a finished product, but that’s OK, you post-capture process them in LR or whatever to bring up the luminance and contrast. But colour should be about right - i.e. reasonably close to the colour embedded in the slide itself.
Just playing very quickly with your combination jpeg, I created a virtual copy in LR and then cropped each so that I just had the LH & RH images. For the LH image taking the WB off the shaded area of the page that the little girl is cutting puts it in a much better place colourwise, and this is just with a jpeg. So for some reason taking the WB off the film rebate doesn’t seem to be working.
Trying to WB off the rebate of transparency film should not be expected to work, and it’s not part of any normal processing sequence I’ve ever seen for digitizing slides, so no surprise here. With proper profiling the colours should be close to correct relative to the media with little tweaking.
Many thanks for your inputs and thinking along.
@Harry : yes, the target looks perfect on my monitor, but it’s a reflective target, not a transparency. Interesting that you were able to get closer by just measuring the white balance off the pic. I did try that in other pics but it still didn’t seem to get me where I needed to be. Will try more if white balance is indeed the culprit.
@Mark_Segal : yes, my calibration could be done better with a transparent target, but as it is, I’m getting something close to the generic Adobe profile, so any differences would be minor with even better calibration, I.e. not explaining the large error I see in my LH picture. Could you please explain why taking the white balance off the end of the transparency roll would be wrong? Don’t I measure what should be white in that way? Sorry, I’m not arguing here but just trying to understand.
Whether the rebate is black or white, those two extremes are poor for doing a colour balance. The purpose of reading the rebate with negative film is to neutralize the orange mask, but we aren’t dealing with negative film or orange masks here. And what’s more, you must use a quality TRANSPARENT target to make a proper profile for “scanning” transparency film. Don’t expect good results if you aren’t making a profile using the materials designed for the purpose and context.
When you said
No, you’re absolutely right of course, the word ‘blank’ set me off in entirely the wrong direction, the unexposed rebate certainly wouldn’t work. In fact I think the OP has used the exposed end of the film, so clear apart from the film base colour, like an extreme highlight. I’m surprised it was so far out in fact.
Sorry for the confusion. Indeed, I used the exposed end of the film, which was blank, showing only the light passing through the transparent base layer. I also took a WB off the 18% grey card in the Colorchecker Passport though while shining the entire light string on there, which should be accurate for the light that hits the film during scanning. Will play around with that and report back. On travel for work now so will take a couple of days
Don’t do any of that. Buy a bespoke transparent scanner profiling target and decent profile making software and use them to make a correct device input profile. THen use that profile for “scanning” your slides. Nothing else.
Solved! Here’s my findings, just in case someone else runs into this issue.
What I was doing initially: Let’s calibrate this baby, using the setup above, shining light from my hacked enlarger that I use for DSLR scanning (full description can be found here: Let's see your DSLR film scanning setup! - #179 by ArnoG) onto a X-rite Colorchecker Passport as seen above, using the 18% graycard in the Colorchecker Passport for the (reflective) white balance, and the X-rite LR plugin to create a custom camera profile. I get:
Measuring the fourth grey pad from the left on the bottom I should get about 50% grey according to X-rite and I measure exactly that, using the custom white balance that results in a color temperature of 4,300 K and a “tint” of +36 in LR. My reflective target seems spot-on, so I presume that my system is calibrated using the 4,300 K, +36 tint, and the custom camera profile hence generated. Scanning my Velvia 100 slide with this profile and white balance, however, results in the off-color LH picture above and the middle picture in:
Based on Harry’s finding (thanks!) that taking a white balance off the paper that the girl is holding gives something closer to what it should be, I suspect no longer a profile error, but indeed white balance, and since I use a Solux daylight bulb, indeed switching the WB to daylight (5,500 K and tint +10 in LR) improves things. My Solux bulb (CRI 99+) says 4,700 K on the box:
So I plug in 4,700 K, tint 0 into Nikon’s NX Studio using the Nikon proprietary “camera neutral” profile and get the picture on the right, which seems very close to my slide.
Using LR, with WB = 4,700 K, tint 0, and my reflective target custom profile things seem too reddish and bold, but using Adobe’s version of Nikon’s camera neutral profile, or Adobe’s own color neutral profile for the camera, results in the pic on the left with only a slight exposure adjustment (EI +0.75), which is very similar to what Nikon’s NX Studio produces (right picture), and similar to what is on my slide. Which profile is used (Negativelab 2.3 neutral, Adobe Color neutral, or Adobe’s neutral version of the Nikon camera profile) only has a minor effect on the color balance.
Lesson learned: A reflective target calibration cannot be used for transparent media scanning, which is what Mark was stating (thanks!).
Overall, apart from creating a custom calibration using a transparent IT8 target which could improve things further (I might do at some point), I’m pretty close already, and at least the colors are no longer horridly off.
I’m still puzzled as to why a reflective target calibration cannot be used for transparent media, but alas, my setup is producing proper results now, pending a further custom calibration using a transparent target for fine-tuning.
Thanks all for the useful inputs and help.
Arno
The answer to your puzzle is that the character of reflected light is very different from the character of transmitted light and your media requires transmitted light for digitizing. Profiles are meant to describe device behaviour under the same conditions for both making the profile and using it. If you make a profile with one kind of light and then use another kind of light for digitizing the media, the profile is no longer providing a correct map of device behaviour in its conditions of usage.
You are fortunate to have put together a kludge that kinda works, but would be on firmer territory implementing principles of colour management correctly.
I appreciate being able to contribute to this forum. I dont have the depth of knowledge and expertise of others - would like to offer:
It is extraordinarily difficult to nail variables on subjective measures - and so each stage of any complex process benefits from being isolated and calibrated - IF that is considered to be of value. I dont see many (i might have missed) references to appa like chromix colourthink for colourspace /volume understanding or the profiles including Chrome-spaces by Jospeph Holmes. As im sure many know,
Photoshop is not linear and managing colour and colour spaces - is for many a horrid journey in that arena - and 3D Lut etc provides effortless adjustment opportunities - both chromaticity and luminosity. Thats all nice - but having the workflow tech from capture through to print, nailed and calibrated. As painful as that can be, at least then there are reference points.
To totally cancel all the above out - William
Eggleston, Dr Elliot Porter, Gordon Parks and even Garry Winogrand didnt have at their disposal any of the incredible kit we have at our disposal. There was pure judgement and a visual integrity that formed the base of discussion on the image. So I may not have helped here much, other than take up column inches. Best wishes
Hi all,
Getting back to this subject with a new idea. Would like thoughts on whether it makes sense, or is complete nonsense.
Following Mark’s comments above, calibration with a transparent IT8 target is best, but one would need one target for each film type (no?) and pray that the target will properly represent the film type that is in use. Anyway, re-thinking this I came up with the following idea.
I happen to have a x-rite colorchecker passport, which can be used to calibrate a DSLR by taking a picture of the colorchecker, which has 24 patches, and then using a LR plug-in, or stand-alone x-rite camera calibration software, a custom camera profile is generated that can be used in LR.
What if I would take a picture of this colorchecker target, probably preferable with the light I use for scanning, on every slide film that I shoot, and then use the scan of that picture as the calibration for that film? Wouldn’t this provide a calibration of the entire scanning string specifically for that film? This solution would be cheaper than buying IT8 targets for each slide film I intend to use, and is possibly also more accurate. This is apart from having a more limited number of patches compared to an IT8 target, but one could of course also shoot a reflective IT8 target as the first pic on each roll, and use that for custom calibration for that roll. I just happen to have the x-rite target for now.
Secondly, I suppose one could also should the target on each roll of color negative film, but I am not sure when to apply the calibration in that case since the pic needs to be converted to a positive first. Neither do I know whether NLP would allow the use of my own camera profile during conversion.
For slides, I presume the above could work wonderfully, unless I’m missing something in my line of thought. For negatives…I’m not sure.
Thoughts?
Arno
Transparent profiling targets are not available for each film type. The norm is one Kodachrome target and one Ektachrome or Fuji target, and that’s about it. At most you need two depending on whether you are digitizing Kodachrome.
I recommend using a bespoke scanner target for profiling a scanner, not a camera profiling target. For the good targets the patch specifications have been carefully compiled and tested for scanners.
Do not attempt to profile negative film (the Hutchcolor website has a lengthy article about this). Use what Nate has built-in to NLP, as it works well.