Online comparision of heritage objects - best colour space?

[cvhmanchester]

I am preparing an online catalogue of C19th prints, providing direct comparison between digital images of the different impressions of the prints (iiif up to 3200 dpi of the original, "true" colour). Depending on the size of and access to the prints, the images are captured with a scanner (Epson 10000XL/Silverfast, 16 colour bit depth) or a camera (OM EM5ii + 30 mm or 60 mm macro/OMWorkspace, ca 12 colour bit depth). Images of the same object created by these 3 different systems render colours very differently when compared side by side on a monitor. And anything I do in Photoshop etc does not make it better. The images are not to be intended to be printed. The underlying database of captured images and metadata is intended to form the basis of an archive as "true" as possible given currently available technology.
A number of questions arise
1. It is suggested that the rendering intent may be one cause of mismatched colours when converting from RAW to files that can be rendered on a monitor. Would Absolute or Relative Colorimetric Intent be better to achieve these aims? This Cambridge COLOR SPACE CONVERSION paper suggests the Absolute method has little usefulness.
2. What is the best way to calibrate these systems so that a colour on the same object creates the same colour value in the image files and thence on a monitor? The absolute value rendered on a specific monitor (a user's standard of colour rendering is not controllable) is of less importance than inter-image consistency.
Thank you

[pschlute]

This sounds like a complex subject, and is certainly not my area of expertise, but I make the following points:

You mention 3200dpi ? if your catalogue is an online one, you need not worry about dpi or ppi. Your images are only made up of pixels. edit: I guess ppi is relevant to a scanner ?

If your three different methods are creating three very different colour renditions it would make sense to stick to one method of capture if possible.

If you are using raw capture, that raw data needs to be processed to produce an image file. Any variation in profiles or any other of the variable settings in a raw converter are likely to give you different results.

Colour-space....for online viewing you should be using sRGB

[William W]

I am preparing an online catalogue of C19th prints, providing direct comparison between digital images of the different impressions of the prints . . . Images of the same object created by these 3 different systems render colours very differently when compared side by side on a monitor.

Hence the question and caveat statement -

What is the best way to calibrate these systems so that a colour on the same object creates the same colour value in the image files and thence on a monitor? The absolute value rendered on a specific monitor (a user's standard of colour rendering is not controllable) is of less importance than inter-image consistency.

Maybe that's not the best question - maybe it would be better to answer: "How can I manage the task, to use only one capture method?"

WW

[cvhmanchester]

Thank you to both of you for your ideas.
1. There are two aims for the project.
First - to capture the images in as high a resolution (spatial and colour) as possible, to provide some (probably limited) future proofing of these archive images. In the 1960s-1990s, newspapers were "preserved" by photographing in B&W on microfilm, now transfered to digital. Many of these tiny images are illegible and many of the orginals were in colour. Then the libraries destroyed the originals. So we need to create a colour (ideally reflectance curve) description of the original that is traceable to international standards measurements. The spatial resolution for image capture is chosen depending on the resolution achieved by the engravers of the metal plates. I have fournd that ca 1200 dpi captures the detail of the engravings from copper plates, but the the best engravers managed detail resolved at ca 3000 dpi from steel plates. This resolution reveals the paper fibre structure, which limited any practical higher resolution for prints on the paper available.
Second -to make the documents available online using the available technology now and into the future. iiif tools on websites enable an image to be zoomed and examined at a huge range of magnifications, like the zoom tool on Photoshop to >100% magnification. The current good colour rendering business standard is sRGB, professional standards are moving towards Adobe RGB, and ProPhotoRGB is on the horizon. To design for present standards is to design in short term obsolescence.
2. Life would be good if only one image capture device could be used. Unfortunately, many of the objects are bigger than A3, the platen size of the scanner, so these have to be imaged using a camera. Using a camera scanning across the object to create a mosaic of images that are then composited is proving unreliable so far. Also the objects are scattered around the world in private and institutional ownership, each of whom has its own image capture kit and software. So we have to provide techniques which can be applied locally.

[DanK]

Given that you are trying to reach iiif standards, I suggest you contact one of the organizations involved in that effort. Perhaps there is someone here who is, but if not, our advice is not what you should be seeking.

[pschlute]

I guess the ppi is relevant to the scanner ?

To future proof your project I guess the best thing to do is keep your raw files secure, and if at some point in the future the web transitions to using a wider gamut like AdobeRGB or ProPhoto, and monitor screens are universally able to display these colours, you can then re-process your images. Until then I would suggest it is best to stick to the most common colour space...sRGB

[revi]

Your main problem seems to be the colour rendition difference between the two/three systems you use.

Could it be useful to create a device profile for each of your systems starting from the same reference card?
Note that for the camera system, such a profile would only be valid for the lighting conditions under which it was produced...

Wrt the rendering intent: that is relevant for conversions between colour spaces, if you have colours in the source space that
cannot be represented in the target space. I'm not sure I'd expect that to happen with images from prints.

[Manfred M]

The solution to me seems that you need to follow a colour managed workflow. That means that all of your capture devices need to be calibrated and profiled to accurately capture colour.

X-rite / Calibrite and DataColor definitely make these for cameras and I believe they may also make these devices to calibrate scanners as well. Each capture device needs to be done separately. You should also scan or photograph a target (at a minimum a neutral gray card, if not a full colour target - again something both X-Rite / Calibrate and Datacolour also produce, ideally every time you scan your target. That way any minor changes in colour quality can be compensated for.

You also need an excellent source of light. A number of people have recommended Solux lights (https://www.solux.net) as been the best out there. LED and fluorescent lights are well known for not having good response curves.

In general, the screens you use to view should be at least 100% sRGB compliant (most cheap computer screens are not), A number of us use wide gamut screens (close to 100% Adobe RGB compliant). These should be set to run at an output of 80 - 120 candela / square meter output and be used in a fairly dimly lit room (under 70 lux at the screen and for critical applications, under 30 lux). Ideally the viewing area will be clear of coloured clutter with neutral coloured (preferably gray) walls.


The two rendering intents we use in photography are Relative Colorimetric and Perceptual. Cambridge in Colour has a good writeup on these and I would suggest you read how they handle out of gamut colour and do some testing to see which one works better for you.

I hope this all makes sense....

[cvhmanchester]

Manfred et al. Thank you all
Yes, I am trying to establish a colour managed workflow, but trying to inter-calibrate the three systems failed with the knowledge I had.
Your points are most valuable, and I have spent the week following them up. So far, I have found 3 relevant colour charts, ISO 12641, Calibrite, Danes-Picta.
You do not mention it, but the best (?) specified calibration chart conforms to ISO 12641 (-1 or -2), which includes a file of batch level measurements for each patch, with individual patch measurements as separate cost item (ca £130). These batch and individual measurements have specified tight variation.
The x-rite/Calibrite calibration chart SG has a calibration file but that does not seem to be updated since 2014, see https://heritagesciencejournal.sprin...al-information. One can get a file with individual patch measurements as separate cost item (£290).
The Danes-Picta charts are much cheaper, but still are also calibrated on a batch level.
I should like to use a single calibration chart for scanner and camera so am still working on which chart to choose and how to implement it into the workflow.
I currently use 150W tungsten filament bulbs, ca 2900 K. Next is to investigate the evenness of lighting and differences caused by lights of different colour temperature, scanner and camera. As you say and I found the hard way, "colour matching" fluorescent and LED lamps don't.

[pschlute]

Beyond my pay grade....good luck