I’ve started scanning some negatives with my Pentax K-5 in the hopes that it will be faster than my Nikon Coolscan V. The Nikon scanner does a great job, but takes a long time.
I scanned my first full roll (Kodak Ektar 100) yesterday and processed it with NLP. Many of the images have a strong cyan cast. It’s most obvious in some photos of snow where the shadows are quite cyan. In other photos, the sky is more cyan than sky blue. I’ve tried correcting rhis in NLP, but haven’t been able to get results I like.
Would need to know more about your procedure for using NLP to be able to help with this. It could be several things. Firstly, at the capture stage with the Pentax, there should be no stray light leaking that could affect the negative - only the light coming through the light source. Secondly, you should set your camera White Point to the same temperature as the light source illuminating the negative. Thirdly, at the processing stage, before you open NLP, in Lightroom you should click the White Balance tool on a blank segment of the film frame that surrounds the image (make sure you capture a bit of this fringe for this purpose). Fourthly, you should crop out everything that surrounds the image itself. THEN finally open NLP and make sure the Convert tab is set for Digital Camera and I recommend Basic as the color model. Then press the CONVERT bar. The colour should now be pretty good “out of the box”, however if you still see a cast, in the Edit tab, I find using LINEAR Gamma and Frontier as the best parameters, then adjust the Tint and Temp sliders opposite to the colour cast you want to remove. This will produce a “CUSTOM” White Balance.
I blocked stray light and scanned at night with all the lights off.
I don’t understand how the camera white point makes any difference given that NLP is processing a raw file. Can you help me understand that?
I clicked the white balance tool on the unexposed part of the negative before opening NLP.
I cropped the image, leaving a very thin border, but used a 2% border buffer and the preview to make sure it wasn’t included in NLP’s analysis.
I opened NLP and selected Digital Camera and Basic as the color model. I selected 1 - Very Low for pre-saturation because in previous tests, the images seemed quite over saturated, and that setting seemed to work best. Next, I pressed CONVERT. On the next screen, I tend to like the Linear tones, but used LAB-Standard for this one. I selected the Natural LUT
Here are the settings I used for the image I uploaded:
I played around with the Tint and Temp sliders, but couldn’t seem to fix the color balance. As I mentioned it was trivial to get a natural looking result for the Nikon scan.
I’ll try it again with LINEAR Gamma and Frontier, and play around with the tint and Temp sliders.
Anyway, I don’t mean to sound defensive, and I do appreciate the advice. I’ve done lots of color correction in a wet darkroom, in Lightroom, and in Photoshop, but am struggling with this.
Re #2, in theory the WB setting on a camera does not affect the raw data, but for this usage only I have tested “Auto” WB against a custom WB that matches the light source and got better results with the latter. I can’t explain more than that, except that perhaps some meta-data re quality of light is being picked-up in the subsequent processing - Nate would need to chime-in on that one.
Re #4, I recommend you leave zero surround.
Re #5, did you try LINEAR-Gamma and Frontier LUT? Just for giggles, try those and see what it does for you, if anything. I’m getting pretty close to what I consider correct WB results with those settings.
The Nikon scan has a magenta-red cast to it on both of my freshly calibrated and matched monitors. To address this, I would first use a hardware and software kit from Datacolor or Calibrite to calibrate and make a custom ICC profile for your monitor.
Snowy, sunlit scenes generally reflect the sky, which is cyan-blue. I suspect the image processed in NLP is the more accurate. Further tweaking of the NLP controls could bring it even closer to neutrality.
To check the accuracy of your scanner or camera/NLP setup, use a standard ColorChecker chart. Photograph it at normal, +1, +2, -1, and -2 stops under direct sunlight. Scan and copy, then see what tweaks are required to match the target as closely as possible. But do so only on a properly-calibrated and profiled monitor, with a correct ICC color management configuration.
I tried zero surround, LINEAR Gamma, and Frontier LUT. I had to raise the brightness quite a bit. It looks a little better, but I still like the slightly edited Nikon scan best.
burkphoto,
I agree that my original Nikon scan has a magenta-red scan, but I think I corrected that in the edited version below it. My monitor (BenQ SW2700PT) is calibrated with a Spyder X Pro, although probably not as recently as yours. I checked the lighter shadows in the snow, and they’re quite neutral based on the RGB values in Lightroom. I agree that shadows on snow reflect the sky, but I would call that blue more than cyan-blue. In other images on the same roll the sky is much more cyan in the NLP scans, but the less cyan Nikon scans look more like real sky to me.
I happen to have photographed a ColorChecker Passport 2 on the same roll of film at -2, -2, 0, +1, and +2 stops from nominal, and I scanned it last night along with the rest of the roll. I have the RGB values for each of the color patches (from http://www.brucelindbloom.com/). I spent some time in Photoshop creating a curve that matched the 6 grayscale values. When I did that, the colored patches looked quite a bit too saturated. I tried to create a curve to match some of the color patches as well, but Photoshop doesn’t like it when the curves get too crazy.
I’ll calibrate my monitor again and spend some time with the ColorChecker photos.
Thanks to both of you for taking the time to reply.
Mark
I calibrated my monitor, then converted the photo of the of the ColorChecker that I had exposed at the nominal metered exposure (Pentax LX, Ektar 100 shot in full sun, by the way). I cropped the digital image to include just the ColorChecker panel with the color and gray scale patches, then ran NLB with Basic Color model, presaturation at 1, LINEAR-Gamma, and Frontier LUT. I set the WB on the 50% gray patch, then adjusted brightness until that patch was as close as I could get to 50% RGB. Having done all that, it looked pretty close to the ColorChecker itself (although it’s always hard to compare reflective surfaces with a computer monitor).
I processed the example photo the same way, and with a few tweaks to the color balance sliders, I’m pretty happy with the result.
Most of the other images on the roll are easier than the snow images, and the ones I’ve done look good with the same settings, but with less color balance tweaking.
I still want to try your suggestion about setting the camera WB based on the light source (Negative Supply 4x5 99 CRI), but I haven’t yet figured out what its color temperature. It still seems like it shouldn’t make any difference to the raw file, but worth a try.
You are welcome Mark. Another question that crossed my mind as I mulled this over yesterday evening is the camera profile (have a look at this thread: Understanding the RAW Camera Profiles in Negative Lab Pro). This is something only Nate can respond to, but I’m wondering whether the camera profile NLP is using for your Pentax K-5 is as good as it should be.
I think the issue here is just that the contrast and saturation are just a bit high in the initial conversion of Negative Lab Pro… so it looks a bit unnatural to our eyes since we know that the shadows in snow should not be that saturated.
Simply lowering the contrast and saturation in Negative Lab Pro will produce a much more natural result here:
You can play around with this yourself in the HSL panel on a positive copy… the shadows themselves are pretty close to the correct hue, they were just too intense.
I think I agree. Here’s what I get with Basic Color model, presaturation at 1, LINEAR-Gamma, Frontier LUT, brightness 25, contrast -5, WB set on mid-tone shadow in snow, saturation 1:
Even with the pre-saturation to 1, the saturation seems high (including my ColorChecker images), and the range of adjustment possible in the next screen is small.
Maybe part of the issue here is that this is Ektar 100 film. When I bought the film I was planning to buy Portra 160, but they were out of stock, so I bought Ektar, which is supposed to have higher contrast and saturation.
In any case, it might be nice to have a wider range of saturation adjustment if that’s possible.
Negative Supply tells me that my 99CRI light source color temperature is 5000K. I copied a negative of my ColorChecker with the camera WB set to 3000K. 5000K, and 7000K, then converted them with identical NLP settings. The results appear to be identical.
Some really insightful discussion and problem solving content in this string of conversation - I can’t help wonder why a calibration tool used (datacolour or calibrate) isn’t used, where subject is a colour checker or similar controlled target. Having that would at least establish whether the film neg image (against the standard) carried bent and or conflicting curves. This would make trouble shooting a whole lot easier. Anyway it seems the root cause is sorted and that’s the main game.
I want to thank you again for your thoughtful and helpful replies to my original post. I found and read your Luminous Landscape Negative Workflow from Capture to Print and Photo PXL Digitizing Negatives with a Camera – Revisited articles, and appreciate the time you took to document your process so clearly and in such detail. I’m sure both articles will be quite helpful to me as I continue to figure out what I’m doing.
Your suspicion about Ektar is likely valid. My experience with it is that I don’t like it… But I worked in the school portrait industry years ago, where we used hundreds of miles of Portra! It’s a much “calmer” film.
I shot a lot of Kodak Supra 400 before between 2001 and 2006 before I switched to a DSLR. I debated at the time about Supra vs. Portra, and, for whatever reason, settled on Supra. Before that I used a lot of 5247 as well as various other films including Kodak Gold. I’ve scanned quite a few of those old negatives with my Nikon scanner. I tried Ektar for the first time a few months ago.
The Nikon scans of (and prints from) the Supra negatives seem to have have a reddish cast, and sometimes I’ve wished I had gone with Portra instead. I’m curious to see how camera scans of the Supra negatives processed with NLP compare to the Nikon scanner and to Ektar.
