School me on color checkers for scanning

[warden]

Hi all,

I avoid color because my color vision isn’t great. Normally that’s not a problem but presently I have a project to do and want good quality color results. I’ll use Portra 160 and an Epson V700 scanner with Vuescan software, and will adjust images using Photoshop. I’ve been seeing products like the one below but have never paid much attention to them:

https://www.bhphotovideo.com/c/product/1649331-REG/calibrite_ccc_colorchecker_classic.html

If you are using them and somehow adjusting scanner or image settings with the aid of a color checker please share your experiences. Thanks!

 

[koraks]

1: Ensure proper viewing conditions in your workstation area; esp. light (high CRI or natural, no direct sunlight)
2: Calibrate monitor
3: Shoot color checker with film/process of your choice
4: Digitize and load into photo editing suite. In scanning, try to make a raw scan and avoid all automatic color and exposure corrections. I find it helps to scan as a color positive since this avoids attempts of the scanning software to deal with the orange mask and inversion. Note that I use Epson scan and not Vuescan; I assume you're at home in the features and settings in the latter and know how to get a (relatively) unadjusted raw scan from it.
5: Adjust curves on image to achieve parity with the physical color checker

No. 5 can be challenging because it's somehow difficult to match a reflective target to something shown on your monitor. In this case, I'd suggest photographing the color checker under good/controlled lighting conditions with a digital camera of decent quality and the white balance set to match the light source used. This yields a digital file that's a little (lot) easier to compare your scans to, as you can place them side by side on your monitor. In your specific case this may be especially helpful since you can do the color comparison with the sampler tool of your photo editing suite instead of relying on a visual inspection. If your color vision isn't too good, such an approach is doomed to fail anyway (sorry to put it bluntly; no offence intended).

Ideally, you would shoot the color checker with multiple exposure settings so you have several under- and overexposed negatives. This ensures that you hit the right color balance and avoid crossover to the extent possible.

Frankly, if accurate color reproduction is a prime requirement, I'd strongly consider shooting digital in the first place. Personally, I believe that color negative's merits are not in this specific arena. Yes, it's possible to shoot more or less color-accurate with CN film. But it's unnecessarily challenging given how relatively easy it is with digital. Of course, if you have other requirements that exclude digital, it's a moot point and you'll have to deal with its quirks.

Finally, there are ways to construct an ICC profile for your color negative + scanning workflow, but it will ONLY work as long as you shoot the film under the exact same controlled lighting conditions. You could try a search for "well behaved profile color negative" here on Photrio. There's a thread somewhere that goes into using a number of open source tools to do this. I've gone through the instructions and decided it's only worth it if you're very dedicated/committed and have a huge volume of very uniform negatives to deal with. Under more regular shooting conditions, I wouldn't bother and just rely on the method I outlined above.

Since questions like these very often end up being a kind of x/y problem, I feel obliged to ask the question why you need "good quality color results" and what this means, objectively speaking. This makes it easier to determine what kind of workflow is likely to be successful.

 

[warden]

Since questions like these very often end up being a kind of x/y problem, I feel obliged to ask the question why you need "good quality color results" and what this means, objectively speaking. This makes it easier to determine what kind of workflow is likely to be successful.

Thanks for your response koraks. "Good quality" is ambiguous I know. I want to achieve a more accurate color representation than my norm as a starting point for image processing, and then adjust to suit. I'm exploring monitor calibration and color checker use as a tool to achieve this.

My colorblindness is significant which is one reason I tend to stick to black and white. I can achieve passable results with color if I'm shooting E6 because I have a color reference baked in the slide, but when I use C41 all bets are off and I end up being complimented on my unintentionally artistic choice of a mid-day purple sky, or someone will ask about why people have green skin in my pics. Since this little project will have people (outdoors and indoors) and use C41 I'd like to give myself a head start with color.

Frankly, if accurate color reproduction is a prime requirement, I'd strongly consider shooting digital in the first place. Personally, I believe that color negative's merits are not in this specific arena.

Yeah I get that but my digital camera is a Nikon D70. I'd rather spend a little money and time developing my technique with C41 instead of investing in a camera that I'll rarely use.

Your steps 1-4 sound good, and while I'm still on the lower slopes of the learning curve I have gathered that I would need to make one image with the subject holding a color checker in each lighting situation if I want to use this tool, and then save a color profile to use on all the shots using that lighting.

Step 5 is the tricky part, and if I can believe the marketing materials it may be automated if the software recognizes the color checker in the image and then generates the profile to match.

Have any of you done step 5 (the automated way) and have you been happy with the result? I'm thinking of the step highlighted at the two minute mark in this video:

 

[koraks]

Step 5 is the tricky part, and if I can believe the marketing materials it may be automated if the software recognizes the color checker in the image and then generates the profile to match.

Yes, if you can get this to work (the software thing), it would make things a lot easier. The main reason I'm slightly skeptical about this is that a single exposure of a color checker doesn't protect you against color crossovers. This is because the densities of a color checker are really close together, while a typical daylight scene can easily exceed an 8 stop range. This means that you may end up with a perfect calibration for colors that are close to the 'middle grey' reflectance, while everything in the shadows and highlights is off. Hence my suggestion to bracket exposures and verify the shadow + highlight exposures as well, and adjust the curves as necessary. However, I don't have a good/fool-proof approach for this especially if you find it challenging to match colors visually.

Yeah I get that but my digital camera is a Nikon D70.

Hey, for most purposes, that will be fine! But I get that you prefer to do this in C41.

 

[MattKing]

Just calibrating your monitor will help a lot.
And if you use the target as a reference standard, it can also help as well.

 

[warden]

This is because the densities of a color checker are really close together, while a typical daylight scene can easily exceed an 8 stop range. This means that you may end up with a perfect calibration for colors that are close to the 'middle grey' reflectance, while everything in the shadows and highlights is off.

That's interesting and I hadn't thought of that so thanks. With any luck someone here has tried the Lightroom or Photoshop integration with these software tools and can report their experiences.

 

[Mr Bill]

I avoid color because my color vision isn’t great. Normally that’s not a problem but presently I have a project to do and want good quality color results. I’ll use Portra 160 and an Epson V700 scanner with Vuescan software, and will adjust images using Photoshop.

Would you mind expounding a bit on the extent of your color vision issues? I did read your post mentioning "green skin," but it doesn't really tell me much. For example, how did you know you have issues? Have you had any color vision tests?

If your color vision is really off, then it might not help to "calibrate" colors on your monitor. If that's the case I'd say your options are pretty much limited to either 1) learn how to "work by the numbers" or 2) have someone assist you.

Another question... about your project. You mention that it "will have people." So how significant are the skin tones? Will it be more along the line of portraiture vs crowd shots, or something else?

FWIW I tried looking at your history here for some background information but you seem to have it blocked.

Ps, in your B&H link, the Image of the colorchecker, can you see a difference in all of the color patches? Or do some of them appear identical?

 

[warden]

Would you mind expounding a bit on the extent of your color vision issues? I did read your post mentioning "green skin," but it doesn't really tell me much. For example, how did you know you have issues? Have you had any color vision tests?

If your color vision is really off, then it might not help to "calibrate" colors on your monitor. If that's the case I'd say your options are pretty much limited to either 1) learn how to "work by the numbers" or 2) have someone assist you.

Another question... about your project. You mention that it "will have people." So how significant are the skin tones? Will it be more along the line of portraiture vs crowd shots, or something else?

FWIW I tried looking at your history here for some background information but you seem to have it blocked.

Ps, in your B&H link, the Image of the colorchecker, can you see a difference in all of the color patches? Or do some of them appear identical?

I don’t have specifics about the color vision except for failing the color vision tests commonly given by optometrists, like the ones where there is a field of dots and you ought to be able to see a number. I see dots for the most part. I haven’t bothered with a test for years as my color vision doesn’t affect me aside from photography.

And yes, I can see distinct color differences in that color checker thankfully.

The skin tones will be significant so I hope to get a hand up with a color checker. And I agree, color calibrating my monitor is less interesting to me and I’m leaning toward the color checker alone, and perhaps using a cell phone image as an auxiliary guide too. It would be interesting to have a phone image side by side with a Portra image to check the numbers on the skin tones.

 

[Mr Bill]

And yes, I can see distinct color differences in that color checker thankfully.

Yeah, one of the guys I worked with had a couple pairs, I think, that he said looked the same to him. So he pretty well knew not to trust his visual color judgment. He mostly worked in an engineering technician sort of capacity but did spend time in our QC Dept dealing somewhat with color, but always in a capacity using densitometers or similar instruments.

The test we considered best, for potential color correctors, was the Farnsworth-Munsell (I forget the number) Hue test. We probably had near a hundred people on file (with double tests). This revealed a great amount of info about color weaknesses. Essentially it's a number of color "caps" consisting of sorta pastel colors in a full range of hues. The test is to try to put them all in the proper order. If you had a weakness in a certain color range you might find 4 or 5 caps in a row where you could not tell them apart. The way it was scored was essentially to make a line graph wrapped in a circle. For each cap in the correct position the score is zero. But for each error the score was how far it is out of place. So a perfect score (fairly rare) was a circle. When color errors occurred the line would have spikes outward from the circle. So it's easy to see, graphically, where someone has issues, and how significant these were. Then, as a rule, if someone has significant spikes in one area then they generally have similar on the opposite side of the circle. Lots of people don't have "fine" discrimination (a lot of small spikes, a jiggly line, all around). But no specific color blindness issue. Anyway, without such a test it's really hard to judge the significance of a color vision issue.

FWIW there is an online version of this test, but I would consider it more of a play toy. Results can be VERY dependent on the specific monitor used. I tried on a few over the years... results ranged from pretty miserable to perfect. On the real test I could, if I spent a lot of time reviewing, get a perfect score most of the time. But in the normal test manner I would typically get 2 or 3 or 4 pairs interchanged. So I know not to trust the online test too far. But... if you also have a friend, one with good color vision, you might be able to compare your score with theirs. It might confirm specific color ranges that you have issues with. And possibly colors that you are OK with. I dunno. Fwiw the color blindness thing is mainly a sex thing, the x-y chromosome thing. Something like 5-10% of males have some sort of color vision deficiency whereas only a tiny proportion of females do. So if you have a lady friend it is almost certain she will not have a specific color vision deficiency. (She MAY have sloppy discrimination, though, just not a specific color issue.)

 

[Mr Bill]

The skin tones will be significant so I hope to get a hand up with a color checker. And I agree, color calibrating my monitor is less interesting to me and I’m leaning toward the color checker alone, and perhaps using a cell phone image as an auxiliary guide too. It would be interesting to have a phone image side by side with a Portra image to check the numbers on the skin tones.

I'm not sure the ColorChecker will help you that much. What I would suggest, given that you seem to be a committed photographer, is to bite the bullet and learn how to work "by the numbers," at least to some extent. To me, the way to do this is to learn how to deal with CIELAB numbers. (Forget about RGB numbers; they mean different things depending on the color space/ICC profile in use.) In CIELAB (aka L*a*b*, or on the internet typically just LAB) the numbers will always keep the same meaning. (They are based on a D50 light source, as i recall, which is loosely the same as daylight with a color temperature of about 5000 K.) Plus they vary mostly in agreement with human perception.

CIELAB is probably not easy for a color photographer to jump to. You were probably brought up on the idea of red, green, and blue being mixed. In CIELAB you have to deal with two sets of "opposing" colors. All of the color info is in the a* and b* values. a* is essentially red vs green, and b* is yellow vs blue, with zero value being neutral. The L* value is loosely brightness - zero is black, and 100 (usually) is equivalent to a perfect white reflector.

When you deal with daylight flesh tones they're gonna fall in a fairly narrow range of a*, b* values. Typically about 18 to 20 with equal values (these are roughly what the ColorChecker flesh patches will come in at). Higher numbers are a more saturated color, lower numbers are less saturated. If someone has a ruddy complexion the a* value will be slightly higher with respect to b*. Or if a yellowish complexion then the b* value will be slightly lower with respect to a*.

You mentioned the possibility of greenish skin. This will be the result of a* being low with respect to b*. For example, say you have typical forehead flesh values of L*, a*, b* = 75, 18, 18. If you wanna turn the skin slightly greenish (seen more as a lack of red) change the numbers to roughly 75, 12, 18. For even more try perhaps 75, 6, 18; this ought to give a good sickly green skin appearance (although green is not actually present, only a lack of red). For real actual green the a* value must go negative, for example L*, a*, b* = 75, -6, 18. (This is actual green mixed in with a good amount of yellow.)

If you are willing to spend the time to learn how these numbers work you shouldn't have any more trouble recognizing bad skin tones. In case you don't know how to get CIELAB numbers... you mentioned that you use Photoshop. As I recall, in the window that shows pixel values you have two readouts. I would typically leave one in RGB, then set the other to LAB. (I presume that normal Photoshop still does this.)

One disclaimer: CIELAB is based on a more or less normal lighting situation (as I said I think a D50 illuminant). If you are trying to do something different, perhaps to give a yellowish tone to suggest a sunset or colored lights at a concert then these aim values don't apply.

FWIW I've spent years and years dealing with color problems in photo labs, and have made hundreds of ICC profiles for digital cameras. Hundreds and hundreds of printer profiles. (In those days i/we used the more expensive Macbeth SG target, something like 140 color patches, as well as its predecessor.) So I'm speaking from some experience when I suggest taking up CIELAB; it's what I would do. But... it's a considerable learning curve so it's not for everybody. It's just a matter of what it's worth for a specific person. Most serious photographers (without a lab guy) would probably profile their monitors (using a hardware puck/software) and work visually. So they don't need to learn numbers. But if your color vision is an issue... well it might be worthwhile. I dunno.

 

[Steven Lee]

@warden To build a color profile for a film scanner you cannot use a reflecting target like the one in your link. You need a transmission target similar to what Lasersoft sells for their Silverfast, or Wolf Faust which works for variety of profiling apps. I have only done this in Silverfast which was dead easy: just follow the prompts in the UI, but the drawback is that those profiles aren't useful outside of Silverfast. Wolf Faust targets come with reference files that can be used by variety of tools, but the instructions will vary based on which tool you're using. But you can start here.

 

[koraks]

@Steven Lee those IT8 targets are nice, but AFAIK the transparencies are only positive slides. No color negative. So they're useful to create a profile for the scanner as such, but that won't help much (at all) with color matching color negative scans.

 

[bernard_L]

@warden To build a color profile for a film scanner you cannot use a reflecting target like the one in your link. You need a transmission target similar to what Lasersoft sells for their Silverfast, or Wolf Faust which works for variety of profiling apps. I have only done this in Silverfast which was dead easy: just follow the prompts in the UI, but the drawback is that those profiles aren't useful outside of Silverfast. Wolf Faust targets come with reference files that can be used by variety of tools, but the instructions will vary based on which tool you're using. But you can start here.

A commercial lab might want to calibrate their scanner. But for an amateur/enthusiast, what matters is the end-to-end calibration from scene to (technically) correct file (before any creative adjustments). [Film -> Dev -> Scanner]

What I've done regularly is shoot on each film a Wolf Faust IT8 reflective target. Ideally sunny weather, sun at 45°, avoid specular reflection. I let vuescan do the inversion but otherwise do no processing and save as 48-bit tiff with data as device colors. The result is a darkish, low-contrast image, because a typical color negative uses only a fraction of the scanner's dynamic range. Then let Argyll use the image of the IT8 to determine a profile, assign that profile to each file, change the profile (and file data) to something standard as AdobeRGB. And only then set the black and white points. Many details are missing; this is not meant to be a how-to. This method takes care of color balance, but also the correct gamma, and the cross-talk between RGB, as evidenced by off-diagonal terms when doing a gamma-matrix profile. I can't see how these complexities can be taken care of if adjusting "by ear" in front of a screen.

What if some frames on that film are taken under different lighting?
(a) Did you decide whether you want to calibrate the scene or the objects?
(b) This was not considered a problem with slide film
(c) Feel free to perform a separate calibration

 

[warden]

Guys I really appreciate all of your suggestions.

The test we considered best, for potential color correctors, was the Farnsworth-Munsell (I forget the number) Hue test. We probably had near a hundred people on file (with double tests). This revealed a great amount of info about color weaknesses.

I took one of those once, but it was the online variety and as you suggest it didn't fill me with confidence for the reasons you state. But the results were humorous!

If you are willing to spend the time to learn how these numbers work you shouldn't have any more trouble recognizing bad skin tones. In case you don't know how to get CIELAB numbers... you mentioned that you use Photoshop. As I recall, in the window that shows pixel values you have two readouts. I would typically leave one in RGB, then set the other to LAB. (I presume that normal Photoshop still does this.)

I'll do some investigating there as I've never paid attention to LAB. Here's a random image from the internet that I sampled in PS with LAB and RGB:

I'll try rescanning a color negative next week and see if LAB gets me closer than RGB. Thanks for the suggestion! If something interesting happens there I'll share the results here.

@warden To build a color profile for a film scanner you cannot use a reflecting target like the one in your link.

Actually I wasn't planning on building a profile for the scanner. My thought was that a "flat" scan would likely be good enough (although I could be wrong) and then Photoshop could use the image with the color checker sample to build a profile for the images with the same lighting conditions. Edited to add: my assumption is that the color profile would be useful for only that one day, and only with the images that share the same lighting conditions. Much like the video I shared where the photographer was correcting the image itself rather than correcting his digital camera.

 

[Mr Bill]

I'll do some investigating there as I've never paid attention to LAB. Here's a random image from the internet that I sampled in PS with LAB and RGB:

Hi, yes, on the forehead (sample 1) the Lab values have a* = 14, b* = 15. The two being similar suggests it's a typical (good) sort of skin tone. But...it's a bit lower than my suggested 18,18 to 20,20, meaning that the color saturation is a bit weak. There are two obvious ways to boost that skin tone color saturation. First is to simply increase saturation in Photoshop. Second is to raise midtone contrast a little; perhaps put a slight s-curve using curves tool. But... since that image already has a wide brightness range in the skin tones an overall contrast boost is probably not the best method.

If you look at the second point, on the arm, it has LAB = 77, 10, 13. Again the a* and b* values are fairly close together, meaning that the hue is not too bad (although the slightly lower a* means a slight lack of red). But... since both numbers are on the low side this means that the color saturation is even lower. Now another issue shows up; the RGB values are higher (red = 217). Since the max possible value is 255 there is not as much headroom to work with (and the lighter skin tones will have even less saturation.). So this is something else to keep in mind when you make adjustments - if you make things too light you won't be able to keep the color saturation up.

Anyway good first exposure to LAB color. Best of luck with it.

 

[koraks]

I'll try rescanning a color negative next week and see if LAB gets me closer than RGB.

The main advantage of Lab is that the ab* remains constant even if L changes. In RGB, all three dimensions vary with luminosity. So a+b captures hue, L captures luminosity. This can help match hues even if their darker or lighter. Not that this is fail-safe, since hues in many cases do shift in reality as they change in luminosity (e.g. cool colors in shadows, warmer in highlights etc.)

My thought was that a "flat" scan would likely be good enough

It will; the matching would be on the end points as @bernard_L proposes. As long as what the scanner does remains constant, it doesn't have to be calibrated to anything. Unless you want to directly scan positives (prints, transparencies), of course. Then it makes perfect sense to profile using an IT8.

 

[Steven Lee]

@warden I see. So your goal is life-like color reproduction, throwing away the inherent properties of film. That's a cool idea, so I would just follow the instructions that come with the target (the supplied software, to be precise) except use a film camera instead of a digital one. Now I want to try this myself too!

 

[Mr Bill]

The main advantage of Lab is that the ab* remains constant even if L changes.

Hi, I would say that for unrestricted Lab this is probably (I think) true. But... when limited by the reality of photographic prints, or even the common RGB color spaces, there are color gamut limitations. In these cases the Lab values will track with limits of the color space.

To demonstrate in a simple way consider that a typical color print brightness is limited by the "white" paper base. To make a print as bright as possible it has to be white. If you want some color to appear then you have to add a colorant which necessarily subtracts some of the reflected light. To make said color stronger more light (of some color) has to be subtracted. So it seems apparent that a strong (saturated) color is only possible when the print is considerably darker. So it seems obvious that as the print approaches maximum brightness, white, that the Lab a* and b* values are gonna have to go close to zero. So some given non-zero a*, b* values, representing a color, cannot be maintained at a very high brightness level, or high L* value. A similar thing happens at very dark parts of the print.

This has always been an inherent limitation of photographic prints - when one starts getting specular reflections from skin, and we wanna print these as near-white, it is not possible to keep the skin color as saturated. Likewise the a*, b* values cannot stay as high as we might otherwise like. If for some reason it is necessary to keep the skin color strong the only obvious fix is to simply darken the print. The same issues exist in sRGB and Adobe RGB color spaces, fwiw.

So these media limitations can also limit the ability of Lab's a*, b* values to stay constant.

 

[warden]

@warden I see. So your goal is life-like color reproduction, throwing away the inherent properties of film. That's a cool idea, so I would just follow the instructions that come with the target (the supplied software, to be precise) except use a film camera instead of a digital one. Now I want to try this myself too!

Well, I don't want to throw away the inherent properties of the film, I just want a little help in avoiding skin tones that are obviously wrong to everyone but me. Hopefully my efforts to remove unwanted color mistakes don't strip Portra of its identity, which would not be cool with me. I do like the Portra products.

 

[bernard_L]

Yes, if you can get this to work (the software thing), it would make things a lot easier. The main reason I'm slightly skeptical about this is that a single exposure of a color checker doesn't protect you against color crossovers. This is because the densities of a color checker are really close together, while a typical daylight scene can easily exceed an 8 stop range. This means that you may end up with a perfect calibration for colors that are close to the 'middle grey' reflectance, while everything in the shadows and highlights is off. Hence my suggestion to bracket exposures and verify the shadow + highlight exposures as well, and adjust the curves as necessary. However, I don't have a good/fool-proof approach for this especially if you find it challenging to match colors visually.

Going back a few hours. While this was written about a workflow where final matching is done "visually" (while I proposed to rely on Argyll to do the matching) this is nevertheless a valid objection to my proposal in post #13. At one point I tried to address this issue by capturing the IT8 target over- and under-exposed, generating XYZ values for each patch over- or under-exposed, and combining the three tables of measured device colors for the the three exposures, as well as the three tables of XYZ values into one virtual, wide dynamic target. Too complex, did not work; either I made a mistake somewhere or there is a fundamental flaw.

Anyway, in practice, the method I propose in #13, even though it suffers in principle from the limitation pointed out by @koraks, works in practice. Part of the reason might be as follows. Very dark values are close to black, regardless; and in a typical scene, very light values are neutral white (I sometimes use a nice cumulus for white point when no IT8 is available) which is the basis for the empirical "auto color balance".

 

[koraks]

@bernard_L as long as the film and scan are reasonably free of crossover, your approach should yield fine results for all intents and purposes. And it reduces the reliance on a visual/subjective evaluation, which even with good color vision is an aspect of uncertainty.

 

[Mr Bill]

Well, I don't want to throw away the inherent properties of the film, I just want a little help in avoiding skin tones that are obviously wrong to everyone but me. Hopefully my efforts to remove unwanted color mistakes don't strip Portra of its identity, which would not be cool with me. I do like the Portra products.

Don't underestimate the color accuracy of Portra. My experience, which has been limited to Portra 160 of over 15 years ago now, and under studio lighting conditions, has shown a pretty good color "accuracy." That was back in the days of optical printing onto professional RA4 paper.

When we did extensive testing of the paper/film combinations we would set up a studio using a variety of models with different complexions and hair colors. Professional electronic flash used, which is roughly what the film is designed for, color temp wise. Also including a variety of colored fabrics along with measurable color charts. We ran these over a wide exposure range. Prints were all hand balanced by professional color correctors and evaluated in "proper" color booths. (We picked the "normal" exposure as a master for best color balance, then every other exposure was matched within about 1 cc color on the flesh tone highlight areas.

When these prints were laid out in a color booth we could lay all the fabric samples on top of the print and they would always be a very close match. The real fabrics differed in that they had slight folds, and therefore shadows, which the prints did not. But if one asked, would you adjust color of the print to better match the fabric, it would barely be worth doing. They were that close. Some of the fabrics we used were strong colors, including bright red,green, and blue, as well as some pastel shades. And these matches held up from film that was roughly a stop underexposed to about 3 or 4 stops overexposed. A professional color corrector essentially could not tell the difference between these exposures by observing matched prints.

Now, the main reason we included the test fabrics was: our business was company-owned portrait studios across the US (and Canada). Back in those days it was common, for example, for families to come in, or perhaps just the children, in carefully selected outfits for certain holidays (Easter was a big one). So it was important that the purchased prints made a good representation of the colors. And they mostly did (there was a 100% satisfaction guarantee).

As I've said before I do NOT have significant in-depth experience with Portra outside of those conditions so I avoid making opinions on those situations. Outdoor photography has more varied lighting which can include bluish light (from the blue sky) filling in the shadows, etc., as well as wide variations in "color temperature" and high UV content. (Professional electronic flash has UV filtration built in, as well as a "color temperature" closely matched to the film aim point.)

 

[warden]

Update:

The color checker thing was a bust. The software and documentation that makes it work are woefully out of date (which means non-functional out of the box) so I sent it back. Ah well.

I did an experiment though that I'll share here. I exposed a roll of Portra160 and had it professionally developed. After each shot I also made an image using my iPhone. I scanned the film using Epson Scan/V700, 48 bit color positive with no color correction. Using Photoshop I inverted the image, and found the white, black and grey points. The images looked close to what I remembered the colors to be, but I then superimposed my iPhone pics of the same images. The inset images are from the phone.

(I'm not sure what happened with the House of Prayer image above, but the film image is way off.

 

[bernard_L]

and found the white, black and grey points

Can you elaborate? Did you find in the scene objects that you thought/knew were neutral grey, neutral white?

Or did you shoot one frame with a proper white/grey/black card and used that as a reference --90% of the effectiveness of a ColorChecker for 10% of the cost. ??

 

[koraks]

@warden that clearly didn't pan out very well. I'd start by applying the same scan settings and curve adjustment to all frames. You've now attempted to color match on a frame by frame basis and thus, you end up with severe color casts depending on subject matter (the House of Prayer, e.g.) It's best to scan all frames in one go using the 'film area guide' setting - which I hope is still available on the V700 series.

with no color correction

Just to be sure, this is the setting I use for this purpose on my old 4990:

Note that by default, the option is set to "Color Control".
The adjustment controls in the main dialog window should be all disabled like so:

If these options are still enabled/accessible, the scanner will still apply automatic color correction to an extent.

My workflow is to take the settings as shown above, scan all frames using the 'film area guide' setting - I usually need to do this in two scans to get the entire roll scanned, but since the color controls are disabled, the output will still be consistent. This is vital!

I then color balance one or two representative frames by whatever means I've decided should work for the purpose. Usually I do something like this: https://tinker.koraks.nl/photography/flipped-doing-color-negative-inversions-manually/
Once I'm happy with the curve adjustment, I apply the exact same curve adjustment to all frames. This will get me close or right on the money, cloror-wise, for the entire roll. Final adjustments can be made to taste on individual frames, but at least this gives me a solid basis.

Note how your iPhone shots have a whole slew of digital manipulation applied to them, including brightness-dependent contrast adjustments, artificial warming of colors especially in low-brightness areas etc. This is the reason why they look so nice, but "nice" and "natural" aren't the same thing. I assume that in your CN scanning exercise you first want to establish a baseline that's as close to the real thing as possible. An iPhone is a poor benchmark for this; I'd sooner rely on a dSLR in RAW mode - even that will skew the results, but it'll be a lot closer to an accurate representation of reality.