How is an IT8 Target manufactured and how does it work?

[guyburns]

I've spent a few hours researching this on the net, but have been unable to find an answer to a two-part question: How is an IT8 target made and what is it showing?

The document called TECHINFO.pdf from FTP.Kodak.com/GASTDS/Q60DATA gives a detailed explanation of what Ektachrome IT8 targets are, but doesn't say how they are made. I want to know whether the patches that appear on a slide target, which are defined in the PDF, have the actual colour values as specified on pages 1-2 of the PDF, or whether the slide targets are generated by photographing a printed target. If the latter, that means that what is on the slide target will not be the specified colour, but a representation of the specified colour as the slide captured it.

To make clear what I am asking, an image of a target (called Generic IT8 Target) is shown near the top of this page: http://www.computer-darkroom.com/it8cal/it8_page_1.htm

QUES about CYAN
The PDF says that:

Columns 13-19 contain cyan, magenta, yellow, neutral, red, green and blue scales which are characteristic of the target dye set. The neutral scale is defined to begin at neutral Dmin and end at neutral Dmax in 12 steps with equal lightness (L*) increments.

That statement does not fully define the various shades in each column. To do so you have to go to the top of page 3 where Dmin and Dmax are specified as having L* values of 82 and 2.
So, for example, column 13 shows various shades of cyan. I assume that the topmost shade is cyan with an L* value of 82, and the one at the bottom has L* = 2, and being an Ektachrome slide, both use the particular type of cyan used in Ektachrome.

Q1: If you compared column 13 of an Ektachrome target with column 13 of a Kodachrome target I assume they will be different because of the different cyans used. Is that correct?


QUES about ROW1 COL1
Looking at R1C1 (top left), this is specified in the table at the bottom of page 2 of the PDF as being a colour of:

Hue angle = 16
L1 = 15
C1 = 10

I don't know what those numbers mean, but I assume it is a colour that is precisely defined.

Q2: If you compared R1C1 of an Ektachrome target with R1C1 of a Kodachrome target will they be identical?


Basically, I want to know what you are getting when you scan a slide with a scanner that has been calibrated with an IT8 target: are you getting the slide, or what the slide was trying to capture? This is the scenario:

1. Take a slide of a piece of paper that has a measured Lab value of (1,1,1) under the lighting conditions when photographed.

2. Assume the slide captures that as Lab (2,2,2).

3. Scan the slide, which comes out with a digital value equivalent to Lab (3,3,3).

4. Now apply the IT8 profile to the scanned image. Do you get (1,1,1) or (2,2,2)?

[crames]

I want to know whether the patches that appear on a slide target, which are defined in the PDF, have the actual colour values as specified on pages 1-2 of the PDF, or whether the slide targets are generated by photographing a printed target. If the latter, that means that what is on the slide target will not be the specified colour, but a representation of the specified colour as the slide captured it.

The way I read it, the aim values specified in the PDF are the values that the target makers attempt to achieve with the dyes in the final IT8 slide or print. Not sure how the manufacturers do it, though.

Q1: If you compared column 13 of an Ektachrome target with column 13 of a Kodachrome target I assume they will be different because of the different cyans used. Is that correct?

Yes, because that column is defined for the particular dye set. Some of the other columns are defined as part of a common gamut set (1-3, 5-7, 9-11), which I guess should be close to the same for all emulsions.  Columns 4, 8, and 12 specify maximum chroma for the particular dyes, so I would expect those columns to vary depending on how much the dyes vary in chroma.

Looking at R1C1 (top left), this is specified in the table at the bottom of page 2 of the PDF as being a colour of:

Hue angle = 16
L1 = 15
C1 = 10

I don't know what those numbers mean, but I assume it is a colour that is precisely defined.

Yes, those are precisely defined colors. Those numbers are CIELAB h (hue), L* (lightness), and C* (chroma). Chroma is roughly a measure of the colorfulness. Definitions here:
http://en.wikipedia.org/wiki/Colorfulness

Q2: If you compared R1C1 of an Ektachrome target with R1C1 of a Kodachrome target will they be identical?

Basically, I want to know what you are getting when you scan a slide with a scanner that has been calibrated with an IT8 target: are you getting the slide, or what the slide was trying to capture? This is the scenario:

1. Take a slide of a piece of paper that has a measured Lab value of (1,1,1) under the lighting conditions when photographed.

2. Assume the slide captures that as Lab (2,2,2).

3. Scan the slide, which comes out with a digital value equivalent to Lab (3,3,3).

4. Now apply the IT8 profile to the scanned image. Do you get (1,1,1) or (2,2,2)?

If the profile is good you should get (2,2,2). You are getting the slide, not the original scene.

Cliff

[guyburns]

Thanks Cliff, for the response. Let's assume that what you see on an IT8 slide target are the colours specified, and assuming the only way to get the image onto the slide is to take a photograph of a reflective target, that implies that the reflective target must have been corrected to accommodate the film characteristics. That must be a complicated process.

1. Print a reflective target. Call it RT.
2. Photograph the target onto slide film. Call it ST.
3. Scan ST (call the resulting scan SST) and generate a calibration profile back to the original scene, RT. You can apply that profile to SST to make it look like RT. What we have now is a slide that does not have the correct colours, but which can be profiled to have the correct colours.
4. Go back to step 1 and reprint using the inverse of the profile. i.e. a new reflective target is printed in which the colours are wrong, but when photographed onto slide, the colours should be correct.
5. Check the colours on the resulting slide against the standard. If the colours are not correct, reprint the reflective target, and iterate the process until the colours are correct.

What a carry on! Seems a lot of extra work, and you don't end up with what most people want anyway (which is a calibration back to the original scene). Instead, what you get is calibration back to a slide. I wonder why targets are not offered that enable calibration back to the original scene? A lot easier to produce, I would have thought.

Having said all the above, I must say that I'm highly sus about the effectiveness of targets assuming a good quality scanner such as the Coolscans. I have no evidence for that statement, just a gut feeling. For a mediocre scanner I could understand the use of a target could show obvious improvements, but if you start with a good scanner… There are so many variables, and stages, involved in translating the IT8 specifications into an actual slide, and then into a profile. If IT8 targets from present suppliers were photographed from a reflective target that has the correct colours (as opposed to the slide itself having the correct colours) who would ever know? Most users would actually prefer such a target because in theory it would calibrate back to the original scene.

I have seen a few comparisons on the net of "before and after" profiling. Didn't seem much in it to me, but they were low resolution images so it was hard to tell. I'd like to see some "before and afters" for inclusion in a PDF (to be made freely available) that I'm putting together about scanning. If anyone using a Coolscan V ED, 5000, 9000, or equivalent quality scanner, cares to scan one or two images and then embed the calibration profile, I would be most appreciative if they could gmail them to me at gdburns. I assume that when I receive the image, if I turn off colour management I will see the "before" scan; and that turning on and applying the calibrated profile will give me the "after" scan. One image should be well exposed with a low density range, and the other preferably underexposed and have a wider density range. Scans at 2000 dpi, 8-bit, and sent as 1-2 MB jpegs would be quite suitable.

[crames]

Thanks Cliff, for the response. Let's assume that what you see on an IT8 slide target are the colours specified, and assuming the only way to get the image onto the slide is to take a photograph of a reflective target, that implies that the reflective target must have been corrected to accommodate the film characteristics. That must be a complicated process.

1. Print a reflective target. Call it RT.
2. Photograph the target onto slide film. Call it ST.
3. Scan ST (call the resulting scan SST) and generate a calibration profile back to the original scene, RT. You can apply that profile to SST to make it look like RT. What we have now is a slide that does not have the correct colours, but which can be profiled to have the correct colours.
4. Go back to step 1 and reprint using the inverse of the profile. i.e. a new reflective target is printed in which the colours are wrong, but when photographed onto slide, the colours should be correct.
5. Check the colours on the resulting slide against the standard. If the colours are not correct, reprint the reflective target, and iterate the process until the colours are correct.

What a carry on! Seems a lot of extra work, and you don't end up with what most people want anyway (which is a calibration back to the original scene). Instead, what you get is calibration back to a slide. I wonder why targets are not offered that enable calibration back to the original scene? A lot easier to produce, I would have thought.

I know what you're saying, but I don't think it's necessary to do the iteration, nor do I think the manufacturers do that.

The reason is that the values in the TECHINFO.PDF are only ballpark values to guide the manufacturer in providing a good range and combination of dye densities in the target. It doesn't matter that they don't exactly match the aim values in the PDF document. What matters are the actual measurements of the target slide, which are in the Q60 reference file provided with the target slide. It is the Q60 reference file that is used when the profile is generated. The Q60 reference files can be found at the manufacturers' web sites (Kodak, Silverfast, Wolf Faust). They are text files that can be read with any text editor, and contain with the actual XYZ and Lab values of the patches .

Profiling the slide rather than the original scene is based on the idea that the slide is the photographic end product, so the scan should preserve the appearance of the slide, including the unique imaging characteristics of the film. Otherwise, if profiling to the scene, the different looks of images shot on Velvia and Ektachrome, for example, would be lost.

Having said all the above, I must say that I'm highly sus about the effectiveness of targets assuming a good quality scanner such as the Coolscans. I have no evidence for that statement, just a gut feeling. For a mediocre scanner I could understand the use of a target could show obvious improvements, but if you start with a good scanner… There are so many variables, and stages, involved in translating the IT8 specifications into an actual slide, and then into a profile. If IT8 targets from present suppliers were photographed from a reflective target that has the correct colours (as opposed to the slide itself having the correct colours) who would ever know? Most users would actually prefer such a target because in theory it would calibrate back to the original scene.

I have seen a few comparisons on the net of "before and after" profiling. Didn't seem much in it to me, but they were low resolution images so it was hard to tell.

I think it is worthwhile to profile, although sometimes the results are subtle. It helps to ensure that you get the actual colors of the slide. All you need is an IT8 target to get started. There are good, free programs like scarse that work very well. The quality of the profile you get varies from program to program. I have been using scarse profiles because I think they give better results in the darker tones near Dmax, although I must admit it's been a while since I compared other software.

[Ernst Dinkla]

1. Print a reflective target. Call it RT.

(snip)

If you started from a Color Checker target that uses something like 10 pigments, unique calibrated versions are available, controlled light, white balance on a true neutral white sample (PTFE or better, see Babelcolor) you could do that 1,1,1, more or less for one camera-scanner. Colors in real life are way more complex though. IT8 targets are all based on 3 RGB dyes (reflective and slide, reference numbers identical per target batch) and have even less to do with colors in the real world. I guess that is also why this approach to profile to film characteristics is the common way. For color negatives your approach is sometimes (but clumsy) used. Other factors like inconsistent film development and mask differences caused by that make it unreliable too.


met vriendelijke groeten, Ernst Dinkla

New: Spectral plots of +220 inkjet papers:
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

[Iliah]

> assuming the only way to get the image onto the slide is to take a photograph of a reflective target

Most of the targets I know are produced using film recorders, some are produced using contact prints. I do not think I know of any single commercial target that is manufactured using a camera. In '97 film recorders were in full use at Kodak's.

[guyburns]

I've got some IT8 targets on the way, but I will never have the ability to set up Argyll, Scarse, or Lprof and get them working. It's a pity about Lprof, because it's based on lcms (http://www.littlecms.com/), which seems to me a stellar performer judging by its low errors as indicated by the graphs on this page:

http://www.tkupfer.de/imaging/Scan_Profiling.html.

But I have no idea how the author of those graphs used lcms; it's only a CMS, not a profiler. Maybe someone could explain.

Anyway, I'm left with Coca, which is based on Argyll:

http://www.nla.gov.au/preserve/dohm/coca.html

The download is at the bottom of the page and it is PC only. As I'm on a Mac, I can't play around with Coca at home -- I'd have to go to a friend's place and try it on his PC. Nothing comes up when I searched Luminance Landscape for the term Coca. It appears that Coca has never been discussed here. For me to install it and fool around on a PC might take ages, so I'm hoping someone can download Coca and try it out and let me know if it might be worth using to generate an IT8 target. If so, with scanner and target in hand, and with Coca on a USB stick, I'll venture into the PC world.