Expired color film?

[mtjade2007]

Both my scanners especially in color negative mode will apply dramatic color adjustments that appear to be very non-linear. I never attempted to reverse engineer the algorithms involved, but they are there and exceed an invert-and-apply-default-linear-curve-adjustment.

I agree with you that film scanners are not simple linear curve recorder of the pixels on the film. But they do have the ability to produce fairly decent scans regardless the brand, and the type of the film as long as the film is a normal one. I believe this is because all C-41 films have to be constrained within a C-41 spec for maintaining compatibility. The scanner can not apply exotic algorithms to produce an output. If they did they would produce weird output from time to time from good normal frames of film. This des not happen according to my experience. They produce problematic scans only when the film is not normal

Here's something you might try with any of your scanners: scan some color negatives (ECN2 or C41, doesn't matter) using the color negative setting in your preferred scanning software. Now scan the same frames using the slide/color positive setting. Attempt to adjust the positive scan to obtain the same result as the color negative scan you've just made. Note if the adjustment curves you need to apply involve only linear adjustments, and if the separate color curves need to cross over in order to obtain the same result. If you want to go further, try the same thing with a different film type - for instance do a scan with fresh Kodak film and one with very expired Konica film. Perform the same routine, and note if the exact same curve adjustments will give the results you produce by doing an 'automated' negative scan.

This gives you an impression of what's approximately going on 'under the hood' of your scanner. I suspect the results may surprise you as to how smart your apparently dumb scanner actually is.

I don't think I understand what you suggest here. Negative and transparency films are intrinsically very different in the design of the emulsion. There is no basis to compare the output of a negative scan to a positive scan of a same negative. The algorithms between a negative and a positive scan should be very different despite the same negative frame of film is scanned.

The basic curve adjustment functionality of Photoshop, which to the best of my knowledge has not changed in the past two decades, is far more powerful than what you can do with just a dichroic head. Certain adjustments that take a few clicks in a few seconds in PS take hours or days of work with masking and many test prints. Sometimes you can luck out in the darkroom and get close by relatively simple things such as a little burning and dodging, or a color-adjusted pre- or post-flash exposure. But overall, the possibilities for corrections with just the most basic curves dialog n Photoshop are far more powerful and convenient than what's technically possible with an enlarger.

I did say Photoshop could do things beyond what a wet printing process can do. But it will do that only the user choose to do it. I at one time long time ago tried but never obtained satisfactory result so I gave it up long time ago.

Now figure what it takes for a scanner to do that regardless of what brand or age color negative film you feed it - just place two different brands and vintages color negative strips side by side and notice the difference in orange mask hue and density. Your scanner automatically corrects for this - and that's a lot more involved than it might seem at first glance, as it requires an adjustment of three separate curves and involves adjusting at least one point as well as the slope of each individual curve (and possibly also the curve shape as well as the start + end point).

Are you talking about before the development or before? I believe the orange mask is there purely for the need of wet printing purposes. The color of the mask may be slightly different from film to film but that is blended into the color curves of the film. I don't think film scanners have different unique translation of the pixels for each different colors of the mask. No film scanner that I am aware of is provided with unique ICC profiles for each known type of film. There maybe one default built-in.

If you are talking about before processing I am sure you know it is washed off by the developer.

 

[mtjade2007]

One possible reason people think film scanner does not faithfully yield true colors of the film maybe because a scan is viewed on a computer monitor. The actual wet printed photo is viewed when we look at a print. Every average Joe's computer monitor is most likely uncalibrated. Some are very aged or improperly set up. That will really cause different opinions about the scan.

I often show my scans to my wife and ask her opinions. She most of the time would say it looks good, no problems, etc. But if I send the scans to my daughter What I often get back is too reddish here, too greenish there, it should be gray , on and on. She views the scans on her monitor, not mine.

By the way, when viewing a wet print the lighting could change your view drastically. The color temperature of the viewing light can affect your judgement of the filtration needed for making the print.

 

[koraks]

they do have the ability to produce fairly decent scans regardless the brand, and the type of the film as long as the film is a normal one

Exactly. Since there's considerable variability in C41 film that will still produce (more or less) normal scans, this is evidence of the adjustments that scanners make on the basis of the input image. The scanning software applies an algorithm to the actual image data, which performs curve adjustments on the separate color channels to produce a result that's expected (by the software engineers) to be fairly accurate.

The scanner can not apply exotic algorithms to produce an output.

I guess it depends on how you define exotic, but I'd call it that - non-linear adjustments with different average slopes for different color channels (=crossover). Here's an example of what my 4990 using the Epson Scan software does on a randomly picked set of (expired!) Portra 160 negatives:

This is with auto-exposure enabled so that the scanning software generates something that approximates natural colors. They're not really natural/accurate and would require manual adjustment, but this is also normal for all scanners that attempt to output color-correct scans from C41 film. The lack of an absolute benchmark or reference makes it impossible for scanning software to accurately reconstruct original scene colors based on just the image information.

Note how the curve adjustment above involves:
* Setting the black point
* Setting the white point
* Adjusting the curve slope for each channel
* Applying a non-linear shoulder compensation
* Applying a non-linear lower-midtone compensation
The exact kind of adjustments each scanner (scanning software) does, is specific to the scanner/software. So my Minolta tends to compensate in a totally different manner, outputting totally different (generally much worse) color corrected images.

The algorithms between a negative and a positive scan should be very different despite the same negative frame of film is scanned.

Yes, that's the whole point of that comparison I'm sorry, it's slightly theory-heavy, so it requires some mental effort to understand what's going on. The purpose of that comparison is to establish the kind of correction curve a scanner (and scanning software) applies to color negative data to output a recognizable and more or less correct image. In the Epson example above we could avoid this whole routine because the Epson software conveniently already shows its actual correction curve, but most scanning software hides or obfuscates this information.

For slide film, there will be no algorithm, typically. Since slide film already contains the absolute reference for output colors on the slide itself, the only thing that matters is linearity of the recording device. This is a matter calibration of the scanner and its software. I suppose this is done by comparing the (inherently non-linear) A/D-converted output of the CCD array to a reference target and then working out R, G and B adjustment curves that are embedded into scanner firmware and applied to each signal being acquired. Long story short: there's no algorithm as such that needs to convert the color data from a color slide scan that the computer receives from the scanner. It's already 'good to go', and is (or should be) a linear representation of the original. We can do the same thing with color negative film and obtain this way a linear representation (more or less exact digital copy) of the negative: this is what the negative looks like with no adjustment at all.

If you scan the same color negative with the color negative settings, the scanning software inverts the signal and applies an algorithm to the input to reconstruct a credible color image. An example of such adjustments I've included above for my Epson scanner. If you then take the positive scan we just did above and open the Curves dialog in Photoshop (or GIMP or whatever), you can start making adjustments to the image on each of the color channels. Try to exactly (insofar as possible) mimic the output that the scanning software gave when scanning as a color negative. Once the images look (more or less) the same, you can tell by the curve shapes you've just worked out what the scanner is apparently doing 'under the hood'. In most cases, you will notice that the adjustments are non-linear. The scanner applies more magic than you seem to realize.

Repeating the same routine on different images can furthermore demonstrate that the adjustment is not constant - it's different for every image you feed through the scanner.

The conclusion from such observations is that whatever adjustments the scanner does, is really a set of assumptions based on no absolute reference at all - it's a wild guess of what good colors might look like based on a set of mathematical relationships applied to the image data.

Your experiences suggest that at least for your scanner, this often works out quite well. Kudos to the software engineers who made the scanning software you use. My experience is that the colors scanning software produce are often somewhat close to how they should be, but virtually never accurate. Statistically, it's virtually impossible that they're accurate much of the time, or even in a small percentages of the images.

Automatic color correction in a scanner with actual consistency/real-life reliability is theoretically impossible due to the real-world restrictions scanners have to work with. The only exception is if an ICC profile is available for a specific scanner and a specific film, and only perfectly developed film that has not deteriorated in any way (before or after processing) is fed through it, and only properly exposed scenes are contained on the film shot under nominal lighting conditions (which will be around 5000K or so). In all other instances, the only thing the scanning software can do is take a stab at it.

Are you talking about before the development or before? I believe the orange mask is there purely for the need of wet printing purposes.

After development.
No, the mask is not just there for wet printing. It's there to correct for the inherent shortcomings of the dyes themselves. This excellent website explains it: http://www.brianpritchard.com/why_colour_negative_is_orange.htm
From this, you may pick up that the orange mask is not, as many people think, just a constant color across the entire film. Part of the mask is actually image-dependent and formed during processing. Since it's dye-dependent, the mask is unique to the particular film; it needs to exactly compensate for the shortcomings of these dyes, in this particular configuration of emulsion layers, etc. This is part of the reason why you see variation in the orange mask between film. Differences in processing also induce variation. The end result of this is that if you take two random bits of C41 film, especially from different manufacturers and/or from different eras, there will be quite distinct differences in how the color information is coded in the dye image and its correction mask. Scanners need to compensate for these virtually infinite possibilities. This is just one more argument why (1) a fixed benchmark is simply not present in color negative scanning and (2) ICC profiling is in practice rather limited in its applicability, because the number of profiles required would be quite ridiculous to cover let's say the scanning habits of an amateur who goes through the family negatives of the past 30 years. Your own use case of using expired film is another very significant complication.

I don't think film scanners have different unique translation of the pixels for each different colors of the mask.

No, so they need to mathematically wing it. See above. The consequence is that a scanner will automatically adjust each color negative image separately, and as a result, its output will vary wildly. That it looks natural to you, means two things:
1: The software of your scanner does a relatively good job at WAGging colors
2: You've simply never noticed the inconsistencies, possibly because you were just happy with the results and didn't see a need to dig deeper - which I think is a perfectly fine attitude and one that likely optimizes the pleasure of the whole process.

I do think that the story of your daughter illustrates not just how monitors vary (yes, they do), but also that people vary very significantly in how they perceive color and the degree of exactness with which they are able to discern and analyze hues. To a large part, this can be trained, although some differences remain between individuals. To me, it sounds like your daughter has a more color-critical eye than you and your wife.

No film scanner that I am aware of is provided with unique ICC profiles for each known type of film.

Indeed. And yes, there may be an ICC profile that comes with the scanner, but for color negative film, it will be nearly useless as it will necessarily be one particular kind of film (usually unspecified), processed and exposed under unknown conditions.

If you are talking about before processing I am sure you know it is washed off by the developer.

No color compensation mask washes off of color film. What washes out are sensitizing dyes and anti-halation dyes. These are unrelated to the compensating mask and the color reproduction story above.
In RA4 paper, the stuff that washes out is also not a dye compensating mask, but an anti-halation mask.

 

[mtjade2007]

Exactly. Since there's considerable variability in C41 film that will still produce (more or less) normal scans, this is evidence of the adjustments that scanners make on the basis of the input image. The scanning software applies an algorithm to the actual image data, which performs curve adjustments on the separate color channels to produce a result that's expected (by the software engineers) to be fairly accurate.

I guess it depends on how you define exotic, but I'd call it that - non-linear adjustments with different average slopes for different color channels (=crossover). Here's an example of what my 4990 using the Epson Scan software does on a randomly picked set of (expired!) Portra 160 negatives:
View attachment 343832
This is with auto-exposure enabled so that the scanning software generates something that approximates natural colors. They're not really natural/accurate and would require manual adjustment, but this is also normal for all scanners that attempt to output color-correct scans from C41 film. The lack of an absolute benchmark or reference makes it impossible for scanning software to accurately reconstruct original scene colors based on just the image information.

Note how the curve adjustment above involves:
* Setting the black point
* Setting the white point
* Adjusting the curve slope for each channel
* Applying a non-linear shoulder compensation
* Applying a non-linear lower-midtone compensation
The exact kind of adjustments each scanner (scanning software) does, is specific to the scanner/software. So my Minolta tends to compensate in a totally different manner, outputting totally different (generally much worse) color corrected images.

Good point about film scanner. The film scanner will do all or some of the above adjustments to the scan. If the owner of the scanner desires to do these adjustments again in a post processing for optimization it can be done by Photoshop. Comparing to the traditional wet printing in a color darkroom latter will be much limited to the turning of the dials of the dichroic filters and print on limited selection of RA-4 paper.

The adjustment by the dichroic filters will apply the filtration from the head to the toe of the entire HD curves across the entire image. On the other hand the curve tool by the scanner can apply to one or all color at a time to the shadow, highlight and mid-tone to a selective color.

The difference between film scanning and wet printing is really in how color adjustment is done. It is hard to say which is better or inferior. One is traditional and the other is modern and more capable. Nevertheless, I do not agree that film scanning does not faithfully produce the colors of the film because of being more capable in color adjustments (burps of the scanner) software or hardware.

The film scanner design engineers could have applied similar adjustments like dichroic filters do. I guess the designers chose the more capable approach to implement the adjustments in the programming of the scanner software. But that does not necessarily mean it is inferior and will not reproduce colors faithfully. One has the advantage of actual human visual involvement in the adjustment process and the other uses a limited built-in automation which will result in the requirement of post processing where human intelligence can apply. Perhaps the need of post processing is why it affects what you think of film scanners.

I apologize if the discussion involving film scanning has gone beyond the intended course of this forum.

 

[mtjade2007]

Yes, that's the whole point of that comparison I'm sorry, it's slightly theory-heavy, so it requires some mental effort to understand what's going on. The purpose of that comparison is to establish the kind of correction curve a scanner (and scanning software) applies to color negative data to output a recognizable and more or less correct image. In the Epson example above we could avoid this whole routine because the Epson software conveniently already shows its actual correction curve, but most scanning software hides or obfuscates this information.

I now see your reason for such comparison. However, scanning a negative with the scanner set to scan slides the output is still a reversed image. Did you mean you then reverse the image so it becomes a positive one? Many scanner owners had done this but had concluded such images are not really useful. The comparison you described shows that the scanner did non linear adjustments to obtain the output. But the HD curve of each color in the negative film is never linear. That's why the color adjustment has to be non linear accordingly. I believe this is a drawback of dichroic filtration because it is linear and why scanner software implements the adjustment with more capable and more accurate method you mentioned which is a fundamental set tool of Photoshop. It is true the adjustments are non linear.

The problem with this non linear adjustment is the scanner is never told what the non linear nature is. Besides the HD curves are not fixed. It varies by the lighting and age of the film, as well as the processing varision. The optimization is better left for post processing to optimize the result.

I have worked with 12 to 32 bit A/D converters (built physical PCBs with A/D and D/A converters). They are all linear devices. I am not aware of existence of non-linear A/D converters. CCDs are also linear too as far as I know. Linearty accuracy is another matter though. but they are pretty linear and accurate still.

 

[koraks]

Let's just leave it at this; there seems to be too much going lost in translation or something.

You're happy with your scans, that's what matters.