Creating an i1Profiler custom patch set

[rxchaos]

I'm creating an icc profile for my Epson7900 printer and the paper I use (Hahnemuhle PhotoRag). i1Profiler generates a patch set, and I can set the number of patches to be generated. But, is there some method where I can generate my own patch set from a digital image I have. The purpose is to create a patch set that mostly contains colors in the image. 

[howardm]

I haven't done it but it appears as though you load an existing profile in the Printer Profile Optimization step and there you can have i1P extract colors from an image to add to the patch set and redo the profile

[rxchaos]

I haven't done it but it appears as though you load an existing profile in the Printer Profile Optimization step and there you can have i1P extract colors from an image to add to the patch set and redo the profile

Thank you for pointing me to the Printer Profile Optimization section of the workflow.  The process seemed straightforward.
- select "Optimization" from the workflow selection
- select the profile to optimize
- select "patch from image" for the source for patches
- set the number of patches - max 30 (Don't know why 30)
- print the test chart
- measure the test chart
- save the profile.

One of the tutorial videos on Advanced Profiling at XRite.com covers a little bit of this

[digitaldog]

I haven't done it but it appears as though you load an existing profile in the Printer Profile Optimization step and there you can have i1P extract colors from an image to add to the patch set and redo the profile

Which is a bit different than doing it from the get-go for the first profile, which can be useful. Plus the number of patches extracted is small, only 20 and fixed**.

This can also be done in ColorPort. The first step is get the image resample down so each pixel represents one color patch then build a txt file (I use ColorThink or PatchTool for this). Use nearest neighbor in Photoshop so you get something like this:

This an ECI2002 45x33 pixels (1485 patches).
ColorThink builds a Colorlist for each patch based on what you assign to the numbers. That can be imported into ColorPort.

Update to illustrate what might not be optimal extraction:

[howardm]

Andrew,

Was there supposed to be an image attached to that last post?

[digitaldog]

Was there supposed to be an image attached to that last post?

http://digitaldog.net/files/ECI2002V_CMYK_45x33pixels.tif

[rxchaos]

Which is a bit different than doing it from the get-go for the first profile, which can be useful. Plus the number of patches extracted is small, only 20 and fixed**.

Thank you Rodney.  This was XRite's response to my question...
That figure was chosen as a balance between selecting the principal colors in the image, and simplifying the optimization.
In the testing that our software engineers did, they found little to no advantage in raising the number of selected patches from a TIFF image.

I'm going to try your suggestion of using ColorPort and resampling with Photoshop.

-c

[rxchaos]

Which is a bit different than doing it from the get-go for the first profile, which can be useful. Plus the number of patches extracted is small, only 20 and fixed**.

This can also be done in ColorPort. The first step is get the image resample down so each pixel represents one color patch then build a txt file (I use ColorThink or PatchTool for this). Use nearest neighbor in Photoshop so you get something like this:

This an ECI2002 45x33 pixels (1485 patches).
ColorThink builds a Colorlist for each patch based on what you assign to the numbers. That can be imported into ColorPort.

Update to illustrate what might not be optimal extraction:

I don't think you can extract and export the rgb values from an image using ColorPort. Its probably possible from PatchTool, but the feature is disabled in the trial version.  Could you please post more information on how to extract the values.

Thanks.

-c

[digitaldog]

I don't think you can extract and export the rgb values from an image using ColorPort. Its probably possible from PatchTool, but the feature is disabled in the trial version.  Could you please post more information on how to extract the values.

You need either PatchTool or ColorThink to do this unless there is some free utility to do so that I’m unaware of. From there, save the ColorList as a CGATS file and ColorPort can build a target from that.

[rxchaos]

You need either PatchTool or ColorThink to do this unless there is some free utility to do so that I’m unaware of. From there, save the ColorList as a CGATS file and ColorPort can build a target from that.

Thank you I'll give it a try.  Any preference between PatchTool & ColorThink?

-c

[digitaldog]

Thank you I'll give it a try.  Any preference between PatchTool & ColorThink?

I use ColorThink out of old habits. Just drag and drop onto the app. Then you need to Extract All Colors (or Unique Colors depending on how you’ve built the image itself) AFTER sampling down. Save as List dropdown command saves it as txt, it’s done.

[digitaldog]

This was XRite's response to my question...
That figure was chosen as a balance between selecting the principal colors in the image, and simplifying the optimization.
In the testing that our software engineers did, they found little to no advantage in raising the number of selected patches from a TIFF image.

Ask them to explain the piss-poor job of extraction from the B&W image provided 
To do this ideally, you would want more than just 20 patches and you’d want the patches extracted to be unique, something the product miserably failed to do in my one example.

[aaronchan]

What's the point of building a "optimized profile" for that specific image ONLY?
From day one, this doesn't make sense to me already.
What I'm doing right now is use an excel and create my own list based on facts like skin tone, shadow, hightlights, primary colors, secondary colors, b&w....etc.
I think a good profile should be generated in the very first time.
Optimization, to me, it's just a marketing tools.

aaron

[digitaldog]

What's the point of building a "optimized profile" for that specific image ONLY?

That isn’t the purpose. It is to insure specific areas of color space you wish to target are created and thus measured and accounted for.
In a perfect world, we would measure 16.7 million possible “colors” for the profile. No guessing. The resulting profile would be larger than many images. Otherwise, target color generation can be much wider than anything you wish to target. Take a few thousand measurements of color all over color space, extrapolate to build a profile. Optimize the target, fewer patches are necessary, you can target color gamut of a specific output device. You can include important ‘corporate’ colors in that target. Take gray balance and how well (or poorly) i1P extracted patches that could be used to optimize gray balance.