ColorMunki Photo / Printer Profiles

[Kevin Sholder]

All,

I've profiled my printer and various papers that I use.  My question relates to the option to optimize those profiles by providing an additional image for it to use in updating the profile.  When using an image to update the photo does it update the existing information with more accurate information?  And if it does, then if you use additional images to optimize that profile, does it become better overall or will you quality degrade?

Something like the point of diminishing return, how much optimization of the profile is worth while?  I'm not going crazy, I just don't have the budget for the higher iPro series yet and I'm trying to get the maximum benefit out of what I have.

I have used the following evaluation image to create the color profiles:  http://outbackprint.outbackphoto.com/printinginsights/pi049/essay.html

Then printed it as well as Datacolor test image found at:  http://www.northlight-images.co.uk/downloadable_2/matrixlarge_2.zip

Along with the Granger and Gamut charts found at:  http://www.luminous-landscape.com/essays/test-charts.shtml

And then 3 of my images to evaluate the paper and printer reproduction.

I'm happy with each profile per the paper type I'm using, just curious if I can improve it at all.

Thanks,
Kevin

[digitaldog]

I'm happy with each profile per the paper type I'm using, just curious if I can improve it at all.

Probably not unless you started with a very small patch sample to build the initial profile.

[Kevin Sholder]

OK, thank you for your help!!

Kevin

[langier]

+1

[Scott Martin]

IMO, if you're starting with ~800 patches there's not much to be gained by optimizing afterwards...

[Kevin Sholder]

Scott,

The ColorMunki Photo only does 5 rows of 10 patches in each row.  It is no where near the 800 that the i1 series of products produce and read from.  Hence the question about using additional images or charts to read from for optimization.

Thanks,
Kevin

[Scott Martin]

Hey you asked right? When I was beta testing that product 6 years ago I printed a bunch of sheets on my 60" inkjet printer and read them. Naturally I have a pro oriented approach and found 800 of these patches a little easier to read than 800 patches on the i1Pro at the time.

When I was testing it on a few Noritsu mini lab systems and some particularly poorly behaved, non-linear small dye-sublimation printers I was forced to take a more consumer oriented route and find a different sweet with far less patches. I was shocked at how it did with just 25-50 patches relative to the PMP and MP engines at the time. I don't know what the minimum is today but at one point we could start with a 23 patch target and these particularly non-linear devices lead us to the importance of including more grayscale patches.

You'll have to decide for yourself where that sweet spot it is - the point of diminishing returns. in the end it's pretty subjective. I personally like to take the approach of starting with a good number of patches to begin with and not optimizing later.

[digitaldog]

About the only time I've seen the optimization useful is when using Marc Levine's 2500 patch gray patch target with some Epson printers and even then the improvements while visisble are quite small. I've never seen using an image work for extraction of patches work properly and in fact, for fun, load a 24 patch Matcbeh Color Checker into the module and see what it extracts. At least in the uber expensive and high end i1Profiler, the product is a joke in terms of what colors it 'sucks out' for optimization. Could be a useful feature but X-rite hasn't really invested the engineering time and energy to make it work as it could.

[Scott Martin]

About the only time I've seen the optimization useful is when using Marc Levine's 2500 patch gray patch target with some Epson printers and even then the improvements while visisble are quite small.

But that's with i1Profiler not the ColorMunki he's asking about right? He's trying to get help wight he Munki, and both of our suggestions about starting with as many patches as you can stand and not optimize is a good one, IMO.

[digitaldog]

But that's with i1Profiler not the ColorMunki he's asking about right?

Same engine. IOW, he's not going to see anything useful without going through the effort of measuring 2500 specially designed patches and even then, the differences are small. For Munki, not worth the bother.

[Scott Martin]

Same engine. IOW, he's not going to see anything useful without going through the effort of measuring 2500 specially designed patches and even then, the differences are small. For Munki, not worth the bother.

But you were starting with a target with a ton more patches right? And your i1P starting target probably had lots of gray patches - so it's not really comparable to him starting with maybe 30 patches. I think the constructive approach would be to find the "sweet spot" number of patches to start with so that optimization isn't useful. The patch generators are different but finding a combination that has maximizes the number of grey patches would be smart.

[digitaldog]

But you were starting with a target with a ton more patches right? And your i1P starting target probably had lots of gray patches - so it's not really comparable to him starting with maybe 30 patches. I think the constructive approach would be to find the "sweet spot" number of patches to start with so that optimization isn't useful. The patch generators are different but finding a combination that has maximizes the number of grey patches would be smart.

Again, even with the Munki, which you well know I worked on as well, there isn't anything useful in the optimization unless you build the initial profile with a tiny number of patches (minimum). Use the max number and move on. Forget about adding images, that part of the software is retarded.

Even with 1700 well designed patches, in i1P, 2500 gray patches can produce slight improvements in some instances if one is willing to print and measure.

[Scott Martin]

Well I think the maximum allowed in the final release is 100 patches now so yes Kevin, I'd imagine you'd see a slight improvement if you optimized with a bunch of gray patches. Try it out for yourself and see if you what you think (and let us know what you think!). I haven't found optimizing from color images to be worthwhile. Reading 3 pages of 50 patches really isn't that bad...

[Kevin Sholder]

Guys, I don't want to start a battle here, however the banter is fun to watch as well as educational. (No disrespect meant)

Scott, yes you are correct the ColorMunki, it only produces 50 patches for the first run.  Dry the print for 24 hours, read it, the software generates another 50 patches based on the first 50.  Print that one and let it dry, read it and generate the profile.  It then allows you to optimize that profile based on an image that you select.  Which is where my original question comes from.

This shows what the layout looks like:  http://www.youtube.com/watch?v=agMO1MbZl2o

So that may put a different spin on it since there is no way to produce greater than a total of 150 patches in 3 sets of 50.  Which is why this only costs about $300 versus the $1500+ for the i1 products.  So I'm just trying to do the next I can with what I can work with.

I've done an optimize with both a color image for a color profile and a b/w image for a b/w profile.  I have not printed a color image with the b/w profile, will try that tonight to see if there is a difference.  As for the color, it has already been done, and as you say minor improvement, so further optimization with this setup may not be worthwhile unless I print out and send the other chart out to generate a profile via an outside service, like the ones that both of you provide.

Thanks again for your input, I'm reading both your websites with great interest at this point, nice images as well!!

Kevin

[Scott Martin]

Right, well thanks for bringing us out of our i1Profiler-centric mindset and back to the ColorMunki release version that your using! :-]

I'd recommend going with a 3 target workflow, optimizing with a B&W (not color) image for color profiles. How's that?

[howardm]

You could use the Argyll CMS software to generate & read a different patch set while using the Munki hardware.

[Kevin Sholder]

Scott, so would you use a b/w target like the ones found at:

http://www.northlight-images.co.uk/article_pages/black_and_white_test.html or

http://www.on-sight.com/download/Onsight_BW_Evaluation.zip or

http://www.on-sight.com/download/Onsight_Gray_v8psd.zip

Or is their something else you would look for in a sample image?

Thanks,
Kevin

[Scott Martin]

Scott, so would you use a b/w target like the ones found at:

Well I would experiment and make my own, but yes, I think these are all great places to start. Consider cropping and using a portion of one of these images. You'll want the full tonal range with an emphasis on extreme shadow and highlight detail.

[digitaldog]

Right, well thanks for bringing us out of our i1Profiler-centric mindset and back to the ColorMunki release version that your using! :-]

Doens't matter. Same engine and same results (different options) but the bottom line is this:

1. Start with a very small patch sample from each, optimization might very well help. Minimum patches to most (many) devices is pretty good considering. Depending on the customer, 50+50 is all they would ever want or need.

2. Use a decent number of patches, 50+50 in Munki (which IS optimization at the secondary stage and isn't offered or forced in i1P) and optimization ranges from doing absolutely nothing (less than Avg 1dE using a few thousand patches to test) or makes a very slight improvement that I submit some wouldn't even see! Im sure we both have customers that would want the 2500 patch optimization and those that don't.

YMMY with Munki or i1P. If someone wants to spend the time and media to test it, go for it. We could even take the two profiles and build a dE report for him using ColorAnt or Colorthink.

[Kevin Sholder]

Scott,

Are we using the b/w image to help with the luminosity values overall then?  Or something else?

Kevin