Source for more advanced profile information

[Wayne Fox]

Having recently acquired an Eye-One i0 ( and now its working quite well thanks to this forum), I'm interested in trying to understand a little more in depth some things I've run across as I try to build profiles.

Specifically, in Bill Atkinsons FAQ he states the measurement data can be "evaluated, averaged, selected and cleaned up before making profiles".

I'll admit I'm getting some great profiles, but my curiousity has gotten the better of me.  How do I "evaluate" the raw data?  How do I clean it up?  Does averaging several readings as he does smooth the data sufficiently to warrant the additional effort?

If anyone can point me the in direction of a book or website that offers some insight into this area, it would be much appreciated.

[madmanchan]

Having recently acquired an Eye-One i0 ( and now its working quite well thanks to this forum), I'm interested in trying to understand a little more in depth some things I've run across as I try to build profiles.

Specifically, in Bill Atkinsons FAQ he states the measurement data can be "evaluated, averaged, selected and cleaned up before making profiles".

I'll admit I'm getting some great profiles, but my curiousity has gotten the better of me.  How do I "evaluate" the raw data?  How do I clean it up?  Does averaging several readings as he does smooth the data sufficiently to warrant the additional effort?

If anyone can point me the in direction of a book or website that offers some insight into this area, it would be much appreciated.
[a href=\"index.php?act=findpost&pid=119762\"][{POST_SNAPBACK}][/a]

You can evaluate the raw data in a few ways. One way is to perform a plot using a tool like ColorThink and see where the data points lie. They should be fairly regular. Then there are some basic sanity checks, e.g., no inversions along the gray axis. Let's say you're doing 4096 patches: the 16 x 16 x 16 regular grid. Well, if patch RGB = (0,0,0) is lighter than patch RGB = (17, 17, 17), then there's something bad going on -- usually a problem with the driver setup.

One can do multiple measurements, average the data, and check each measurement against the average to catch outliers -- measurement errors, if you will.

If you look at Bill Atkinson's FAQ, you'll notice that for his "premium profiles" he prints each target multiple times, in different orientations. At first I didn't understand why. Then I did some tests myself and found out. At least with Epson printers, the ink laydown isn't uniform across the sheet, edge to edge (even when one is printing in normal mode, not borderless mode). For example, I printed a row of 20 black patches from left to right across a sheet. 18 of the black patches measured nearly identically. The 2 located closest to the left edge of the sheet measured significantly lighter. I repeated the test multiple times and found the same thing: density isn't uniform across the sheet. Hence Bill's method of printing multiple targets, oriented differently, and averaging the results.

[eronald]

The free Xrite Colorport utility can, I believe, do multiple measurements and average. PMP MeasureTool can probably do the same.

I have heard that averaging is a useful trick when reading canvas or rough materials, or prints made with a very big dot size (eg. billboards).

Although printing testcharts in various orientations etc makes sense for someone doing careful testing, the real issues with profiles used in photography are not going to be at that level if the instrument is an IO. Or at least they shouldn't be.

If you are really needing repeatability, problems with the spectro as it heats up while reading etc etc. will also be relevant.

I don't think it obvious at all that a wide-format prints the same at the beginning and the end of a line, and indeed have heard the Canons do not, as they are thermal printers. Some printers have head redundancies, but these again may create variances in the output.

Do you really want to go there ?

Are your eyes really, really that good ?

Edmund

[digitaldog]

I'll admit I'm getting some great profiles, but my curiousity has gotten the better of me.  How do I "evaluate" the raw data?  How do I clean it up?  Does averaging several readings as he does smooth the data sufficiently to warrant the additional effort?

[a href=\"index.php?act=findpost&pid=119762\"][{POST_SNAPBACK}][/a]

I don't find profile averaging useful with well behaved devices. A press, that's one thing (unless it's a modern, very stable, computer controlled press). But for an average inkjet? Use a decent patch sample like Bill's 4096, use a good instrument that isn't going to measure the patches incorrectly and will do multiple samples (example, the EyeOne Pro measures 100 per second so it is averaging measurements per patch) and be done.

[Scott Martin]

I'll admit I'm getting some great profiles, but my curiousity has gotten the better of me.  How do I "evaluate" the raw data?  How do I clean it up?  Does averaging several readings as he does smooth the data sufficiently to warrant the additional effort? If anyone can point me the in direction of a book or website that offers some insight into this area, it would be much appreciated.
[a href=\"index.php?act=findpost&pid=119762\"][{POST_SNAPBACK}][/a]

There are different tips for profiling different type of processes (inkjet with driver, inkjet with RIP, photo paper processes, laser, offset press, etc). Mention specifically what process you are working with and people can provide tips for that particular process.

If you really want to geek out and learn more about these details you might consider taking a class or get private training as there aren't any books or literature that go into these issues as closely as it sounds like you would like to go. There are so many little things to discuss that it's nearly impossible to be immersed in this process and write about all of it. There is only so much we can talk about without actually looking at prints that speak volumes. Perhaps I could quickly touch on a few things that you might find interesting:

1) As for evaluating a printer's state prior to profiling consider printing an evaluation image (without a profile) that has linear gray and color bars, skin tones, extreme highlight and shadow detail etc. For the profiling process to work well with a generic profiling target a device needs to be able to produce a black to white gradation with even density distribution and without a color cast. If you are working with a device that isn't particularly 'well behaved' like this then you may need to use MonacoProfiler's 2 step profiling method that generates custom profiling targets for a device's specific non-linear state.

2) If you haven't done so already, calibrating a process with an advanced RIP (like GMG ColorProof, EFI ColorProof, ColorBurst, etc) prior to profiling is very enlightening and will help you understand what must happen prior to profiling for the profiling process to work optimally. Most of these RIPs have 3-5 day certification classes that go into these issues with incredible detail.

3) ProfileMakerPro users often overlook the question "Is there another profiling application that would profile this process better than PMP?" Make sure you are using the best software you can before spending too much time massaging a process that may never provide the results you are looking for.  

Scott Martin
www.on-sight.com