How to use a scanner profile?

[Ciaran ODuibhin]

My aging flatbed scanner seemed to be losing its colour accuracy, so I got some profiling software (Lprof), and a calibration chart, and made a profile. My hope was to be able to apply this profile to recently scanned images, to improve their colouring. Does that make sense?

I'm now looking for a way to apply the profile to my jpgs, to produce corrected jpgs. I don't really want anything as complicated as Photoshop or Gimp. Is there something that is straightforward to use for this seemingly straightforward task?

Second-best would be a way to embed the profile in a jpg, so that an application like IrfanView could take account of the profile when loading the jpg, but again I can't see anything simple or free for this purpose.

I would be grateful for any advice.

[Jeff Mottle]

My aging flatbed scanner seemed to be losing its colour accuracy, so I got some profiling software (Lprof), and a calibration chart, and made a profile. My hope was to be able to apply this profile to recently scanned images, to improve their colouring. Does that make sense?

First, creating a profile of your scanner is intended to describe the characteristics of that scanner. This is done so an image generated with that scanner matches (as close as possible) to the original scanned object under a known lighting. It will not improve or regain lost performance of your scanner. If for example your scanner is no longer able to accurately capture really saturated reds, a profile will not bring that back. Garbage in, garbage out.

I can't speak to the accuracy of this open source profiling app (Lprof), however the basic process your describe is correct. Scan a calibration target and then use the software to determine what was scanned vs what the known color patches were on the target to generate a profile. It sounds like maybe you printed this target? If so, I can't see this being very accurate. In fact, if you did, I would stop now. The target has to be a commercial target that was printed to ensure the swatches are of known values.

I'm now looking for a way to apply the profile to my jpgs, to produce corrected jpgs. I don't really want anything as complicated as Photoshop or Gimp. Is there something that is straightforward to use for this seemingly straightforward task?

I've personally never used anything other than Photoshop. It does appear Gimp does support color management, though I have no experience with how they have it implemented. Affinity Color is another app that a lot of people are using if they opt not to use Photoshop. If you're going to get involved with color management, it's kind of assumed you're going to use something more than a free or built in Windows image viewer. If you want to get your hands really dirty, there are some code bases you can use to apply ICC profiles here: http://www.color.org/opensource.xalter

Personally, I would just use PS or the like and be done with it. There are a number of viewers that read embedded color profiles, but you need a full color management engine if you want to do anything more than just apply a profile.


The next question I have, that you did not really specify, is what you are trying to do? What are you scanning, why are you scanning it and what are you trying to do with the output images? This might help me provide some more guidance. While assigning profiles is a pretty simple task that does not require the color management engine to be fully set up, if you are doing anything beyond that, you're going to likely need a fully functional image editing app that supports color management, and have the color management properly set up in that app. You are likely also to require a calibrated and profiled display and perhaps even paper/printer/ink profiles for printing.

[Ciaran ODuibhin]

Thanks for that quick and helpful reply.

What are you scanning, why are you scanning it and what are you trying to do with the output images? This might help me provide some more guidance. While assigning profiles is a pretty simple task that does not require the color management engine to be fully set up, if you are doing anything beyond that, you're going to likely need a fully functional image editing app that supports color management, and have the color management properly set up in that app. You are likely also to require a calibrated and profiled display and perhaps even paper/printer/ink profiles for printing.

I'm not trying to do anything high-powered — just scanning family photographs, and finding (unless it's my imagination!) that recent scans look "duller" than earlier scans when viewed on-screen. And blaming that on a gradual deterioration of my scanner. My only object is to view photos on-screen, or maybe print them, and have them look more like the originals. But I'm not seeking perfection, or consistency across different display devices, with all the complexities involved in that.

First, creating a profile of your scanner is intended to describe the characteristics of that scanner. This is done so an image generated with that scanner matches (as close as possible) to the original scanned object under a known lighting. It will not improve or regain lost performance of your scanner. If for example your scanner is no longer able to accurately capture really saturated reds, a profile will not bring that back. Garbage in, garbage out.

I realize that, if the scanner is producing the same output from different inputs, this can't be remedied later. I was hoping that the "drift" of my scanner wasn't of that type, but would preserve relative differences, while not keeping the absolute values as they were in the original. But I don't know if that's a realistic assumption.

I can't speak to the accuracy of this open source profiling app (Lprof), however the basic process your describe is correct. Scan a calibration target and then use the software to determine what was scanned vs what the known color patches were on the target to generate a profile. It sounds like maybe you printed this target? If so, I can't see this being very accurate. In fact, if you did, I would stop now. The target has to be a commercial target that was printed to ensure the swatches are of known values.

I used a calibration target purchased from http://www.coloraid.de/ . I scanned it , but I haven't done anything with the profile generated from it, except to save the profile in a .icm file, while I try of discover how to make use of it.

I've personally never used anything other than Photoshop. It does appear Gimp does support color management, though I have no experience with how they have it implemented. Affinity Color is another app that a lot of people are using if they opt not to use Photoshop. If you're going to get involved with color management, it's kind of assumed you're going to use something more than a free or built in Windows image viewer. If you want to get your hands really dirty, there are some code bases you can use to apply ICC profiles here: http://www.color.org/opensource.xalter

Personally, I would just use PS or the like and be done with it. There are a number of viewers that read embedded color profiles, but you need a full color management engine if you want to do anything more than just apply a profile.

I hope this makes it clearer what I'm trying to do, and what I'm trying to avoid doing. "Just apply a profile" describes exactly what I want to do — apply the profile I generated to the photos I scanned, to compensate for scanner drift. I realize the photos may still look slightly different on different monitors and printers. Conceptually that seems simple, although it may not be simple behind the scenes, and I'd rather not get involved in a full CMS to do it. But it begins to look like I may give Gimp a go, though I first want to check out the links you gave to http://www.color.org/opensource.xalter

One question: I've seen that some software, including Gimp and Photoshop, claims to be able to insert a profile in a jpg, but why do they not take a further step and produce a new jpg with the effect of the profile taken into account?

Once again, many thanks.

[Jeff Mottle]

...finding (unless it's my imagination!) that recent scans look "duller" than earlier scans when viewed on-screen. And blaming that on a gradual deterioration of my scanner.

It could very well be your scanner, but if you changed your display, display settings or if your display is degrading or defective it could be that too. One way to determine if your scanner is changing is to profile your scanner now and then in say 3-6 months and compare the gamut of the profiles in a gamut viewer. If they are not identical and have large differences, then you'll know what is changing and in what area of the gamut.

My only object is to view photos on-screen, or maybe print them, and have them look more like the originals. But I'm not seeking perfection, or consistency across different display devices, with all the complexities involved in that.

To do this you need a fully color managed workflow.

1) Your scanner has to be profiled
2) Your monitor has to be profiled and calibrated
3) You need to be working in a color calibrated environment (like Photoshop) and have that properly configured
4) You need to use proper printer profiles when printing that match the paper/ink combination. Or if you're using your own paper or ink or if your printer is not intended for professional use, create your own profiles.

None of this is super easy to set up and takes some work and research to understand the mechanics of what you're doing.

I realize that, if the scanner is producing the same output from different inputs, this can't be remedied later. I was hoping that the "drift" of my scanner wasn't of that type, but would preserve relative differences, while not keeping the absolute values as they were in the original. But I don't know if that's a realistic assumption.

A profile can not preserve the characteristics of the scanner to be later used when the scanner further degrades. It ONLY describes the characteristics of the scanner at one moment in time, so that scans done at that time are accurately captured. Devices like printers generally are pretty stable and don't have their characteristics change over time (assuming same paper and ink). Devices like monitors however do change over time and need to be regularly calibrated and re-profiled. So for example, you could not use a monitor profile that is 3 years old and expect it to accurately describe the display at a later date. Generally scanners lean to the more stable side. Does not sound like that is your case though.

I used a calibration target purchased from http://www.coloraid.de/ . I scanned it , but I haven't done anything with the profile generated from it, except to save the profile in a .icm file, while I try of discover how to make use of it.

Ok great. Assuming that profile is generated correctly. You can then open your scanned image in Photoshop and "ASSIGN" the profile to the image. That way it will look a close as possible to the original photograph. There are a number of assumptions here to ensure that happens:

1) Your display is properly profiled and calibrated.
2) Your display is capable of displaying the gamut of colors in the original photograph. This should generally not be an issue provided it's not a super old display, really cheap display or defective.
3) You are viewing the original photograph under the same color temperature of lighting and same luminance of light that your display was calibrate to. For example D50 lighting or 6500K etc. If you have calibrated to 6500K and then view your photo under dim lit fluorescent lighting, your photo will never match the screen. In higher end workflows you would use a light booth that has an adjustable color temperature bulb and a luminance control to match your display luminance. You can also get Solux light bulbs and a desk lamp to get your pretty close too (http://www.solux.net/cgi-bin/tlistor...ges/index.html) or calibrate to 6500K and view in diffuse natural sunlight.

As an aside, if you are going to calibrate your display, you need a proper spectrophotometer or colorimeter. (I prefer Xrite products over Datacolor) Do not try to use those online visual screen calibrators. It's a pointless waste of time.

Conceptually that seems simple, although it may not be simple behind the scenes, and I'd rather not get involved in a full CMS to do it.

What you describe kind of requires it. Color management is not something you can do in parts. It's all or nothing. At the very least you need to profile and calibrate your display. If you don't do the rest, you're just spinning your wheels and will not do any better than just guessing.

One question: I've seen that some software, including Gimp and Photoshop, claims to be able to insert a profile in a jpg, but why do they not take a further step and produce a new jpg with the effect of the profile taken into account?

You ASSIGN a profile to an image when you want to assign meaning to the underlying color number when you know the device that created the image (a camera, a scanner etc.) When you save the photo the profile will be embedded if you opt to save the profile with it. In Photoshop there is a checkbox to save the profile with the image.


There are lots of books out there on color management out there, I've read most of them. If you want the cliff notes for use in our industry, I wrote a 52 page chapter a few years ago for a 3ds Max book. That chapter can be found here: https://www.smashwords.com/books/view/233521 a few things have changed, but much of what you need is there and in more concise format.

Happy to answer any other questions you have though. There are a lot of variables so be prepared if you want to get close, you can't just wing it.

[Ciaran ODuibhin]

Many thanks, Jeff, for all your detailed advice. It is appreciated.

Does anyone know how to get a copy of jpgicc.exe for Windows (Vista)? From what little I've seen about it, it might be useful. There are a couple of downloads on the net, but I'm not convinced they are safe.