I tried a Custom Profile

[bwana]

ok, i get it. there is variability in profile construction. i also appreciate the benefit of reprofiling a new batch of paper (even one you've profiled before) or perhaps a new batch of inks. But perhaps an issue no one has talked about is the number of samples used to construct a profile.

That german provider has 3 'levels' of profiles ? If one made a print from the same printer three times, each with a 'fancier profile' , how much difference could there be? Obviously enough otherwise they wouldn't offer three levels of complexity. (But then again Chrysler made vehicles with the option of Corinthian leather)

It seems 'more is better' so how about actually using the colors in the image you want to print. I know it sounds crazy ( you'd have to construct a profile for each image) but then again, when you consider the actual colors of an image ( the gamut of an image?) the best profile to use would be the one optimized for it. Of course the tech for this does not exist. But it is conceivable. Patches 1 mm square could be printed and you could get  50000 patches on a single letter sized page, and the patches could be generated from the colors of an image. of course you'd need a special spectro for this. or maybe not. Maybe DSLRs in the future will have a spectro mode where a usb cable can directly transfer the measurements from the sensor to the profiler app.

But before that extreme, one might be able to generate profiles for landscapes (lots of blues and greens), profiles for people and portraits (lots of yellows, reds, and oranges.) It would seem a lot of the variability in profiles from one service provider to another might depend on patch selection.

[digitaldog]

Not crazy per se; some products extract colors from images for a patch set. My custom post-optimization target has some. What colors in what color space along with what other patches and where in color space is part of the Sausage Making. Not everyones sausage is created equally.

[Doug Gray]

ok, i get it. there is variability in profile construction. i also appreciate the benefit of reprofiling a new batch of paper (even one you've profiled before) or perhaps a new batch of inks. But perhaps an issue no one has talked about is the number of samples used to construct a profile.

That german provider has 3 'levels' of profiles ? If one made a print from the same printer three times, each with a 'fancier profile' , how much difference could there be? Obviously enough otherwise they wouldn't offer three levels of complexity. (But then again Chrysler made vehicles with the option of Corinthian leather)

Obviously, more patches, the better a profile's Colorimetric tables. But the big variable is the printer. For instance my 9800 has significant "lumps" along the neutral axis and a standard, iSiS 957 single page profile works pretty good for most colors but there are some areas where it's not so good and the neutral axis is a big one. This is important because perception is more sensitive to small changes on the neutral axis. So I've made a special patch set with 4k patches that has a concentrated set of additional, near neutral, patches as well as having I1Profiler add additional patches where it sees areas in the mapping that need improvement.

OTOH, my Pro 1000 has much smoother mapping and the 957 default patch set does a good job on the neutral axis with a limited set of neutral patches.

So how many you need really is a complex function of the printer and there is no way to really tell how much improvement you get with purchased profiles but it's likely the older the printer the more patches would help. But at best it's subtle and I wouldn't be surprised to see no visual print difference at all.

[Alan Goldhammer]

THAT one surely is.

Why do you always jump into a thread to troll someone?  It's getting pretty tiresome.  Stick to sending in image for the home page, you will have more success.

[Rhossydd]

i also appreciate the benefit of reprofiling a new batch of paper (even one you've profiled before) or perhaps a new batch of inks.

For OEM inks and high quality papers, probably not worth doing every time you get new batches. QC is pretty damn good for most major manufacturers.

But perhaps an issue no one has talked about is the number of samples used to construct a profile.

Over the years this has been discussed at very great length, many times.
It's not quite as simple as more=better. An appropriate set for the software is more important for overall results. If you back read through here you'll see mention of certain 'sweet spot' patch sets giving better results in i1Profiler, rather than just 'more is good'. Hit the right numbers and you may see slight improvements in certain tonal areas, but improvements can be subtle and most easily seen when using synthetic test images. (You'll read lots of comments about Bill's balls)

It seems 'more is better' so how about actually using the colors in the image you want to print.

Yes, all possible. You can add extra patches for specific colour ranges, but unless you are actually having issues with particular colours it's probably not worth the expense and effort.
The two specialities that I've come across that benefit from this approach are building profiles for monochrome work and non-Caucasian skin tones.

[Slobodan Blagojevic]

Why do you always jump into a thread to troll someone?...

Only to correct an obvious stupidity, like Andrew’s failed attempt at logic.

[digitaldog]

Only to correct an obvious stupidity, like Andrew’s failed attempt at logic.

The absurd is the last refuge of a pundit without an argument
Or a pundit who hasn't a clue about the topic he's posting in. As the paper trail shows.

[Slobodan Blagojevic]

...  who hasn't a clue about the topic he's posting in....

Then we are even. You don’t have a clue about the logic you are attempting to use.

[Alan Goldhammer]

Only to correct an obvious stupidity, like Andrew’s failed attempt at logic.

Why don't you go back to Page One of this thread and re-read your posts where you clearly were going overboard by pestering the OP regarding his experience with a Custom Profile.  That's not "...corecting an obvious stupidity..." rather it's an crude attempt to rile things up when they did not need to be.  You've done the same thing on numerous other threads.  If you don't have anything to contribute on the technical threads just don't post.

[digitaldog]

Why don't you go back to Page One of this thread and re-read your posts where you clearly were going overboard by pestering the OP regarding his experience with a Custom Profile.  That's not "...corecting an obvious stupidity..." rather it's an crude attempt to rile things up when they did not need to be.  You've done the same thing on numerous other threads.  If you don't have anything to contribute on the technical threads just don't post.

Yeah ask him about his experience building custom profiles what kind of spectrophotometer he has what software he is using. What supreme logic he has for building a perceptual table. 

[bwana]

...
The two specialities that I've come across that benefit from this approach are building profiles for monochrome work and non-Caucasian skin tones.

ICC profile for monochrome?! help me understand why that is even a thing. It would seem the gamut for a B&W image is much smaller than that for color. The tone shifting from computer to printing would only be shades of gray. And unless you have cyborg spectrophotometer eyes, how would you even know. It's not like skin tones, where our brain is hardwired to analyze faces and the slightest abnormality in color is identifiable.

Also, I keep trying to get back on topic but it seems there are some old wounds here.  Can the moderator please step in.

[digitaldog]

ICC profile for monochrome?! help me understand why that is even a thing. It would seem the gamut for a B&W image is much smaller than that for color. The tone shifting from computer to printing would only be shades of gray. And unless you have cyborg spectrophotometer eyes, how would you even know. It's not like skin tones, where our brain is hardwired to analyze faces and the slightest abnormality in color is identifiable.

Also, I keep trying to get back on topic but it seems there are some old wounds here.  Can the moderator please step in.

He told us he's using i1Profiler in the same post you reference about monochrome profiles. That product doesn't produce such profiles. The idea of a profile for skin tones (Caucasian or otherwise) doesn't make any sense at all. Not when you consider what profiles do. Profiles know nothing about color in context! They only know about individual pixel values. Like what you see if you were to open a shot of skin tone in Photoshop and zoom in to 1600%. Not until profile engines work on color appearance modeling and i1P nor any product I know of does.

Colorimetry and the dE testing and the current profile technology is based on color perception. It is not about color appearance. The reason why viewing a print is more valid than measuring it is because measurement is about comparing solid colors. Solid colors have no idea what a skin tone is or should appear like. Color appearance is about evaluating images and color in context which measurement devices can't provide. Colorimetry is about color perception. It is not about color appearance. Colorimetry was never designed as a color appearance model. It's not designed for imagery at all. Colorimetry based on solid colors in very specific ambient and surround conditions. And today, that's how our profiles are created.

[dehnhaide]

Why do you always jump into a thread to troll someone?  It's getting pretty tiresome.  Stick to sending in image for the home page, you will have more success.

+1


Sent from my iPhone using Tapatalk

[nirpat89]

But before that extreme, one might be able to generate profiles for landscapes (lots of blues and greens), profiles for people and portraits (lots of yellows, reds, and oranges.) It would seem a lot of the variability in profiles from one service provider to another might depend on patch selection.

ColorMunki Photo (now called i1Studio) can already do that.  I have successfully used its tweaking capability to make a special profile to more accurately print the shades on my monochromes (warm-toned B&W.)   The original default profile created for the regular color gamut had a distinct greenish cast in the shadows that disappeared with the tweaked profile.  For that alone I would say the cost and effort was worth it.  I don't know if any of the custom services offer this capability or not.  I have not done any tweaked profile for a color image yet.  But I don't see why it would not work.  Presumably one can create a profile for each image to print.  It only takes one more 8x10 print to achieve that.   

:Niranjan.

[Doug Gray]

Colorimetry and the dE testing and the current profile technology is based on color perception. It is not about color appearance.

Perhaps the best demonstration of this can be done in a typical home environment with regular, not proofing, (2700K-4000K) lighting if one has a colorimeter or spectrophotometer that can read out CIE XYZ values.

Place a piece of matte paper on your desk. Measure the XYZ of reflected light off the paper being careful to avoid shadowing the room light so the paper is illuminated evenly.

Now create a large size white area roughly the size of the sheet of paper in Photoshop. Then use Curves and adjust the RG and B until the image measures the same XYZ values you got from measuring the paper on your desk. Go by the numbers, don't even look at the image. This takes a bit of time because XYZ are each effected by each of the R, G and B levels. But after you have made the image color XYZ match the paper readings now look at the paper on the desk then your screen.

They won't even be close. The image on the display will look a dingy, darkish yellow in comparison to the paper on your desk. Colorimetrically, they are the same. That's the Perception part. The difference in how they appear is the Appearance part. The difference is partly context, the different surround, and partly cognitive. That is simply knowing that the white paper is white.

If the room is evenly lit and you can pick the paper up and bring it near the image on the display. when you get within a few inches or so the paper and monitor will, almost magically, shift and suddenly match quite closely.

The effect is so strong it is absolutely mind boggling.

[Rhossydd]

ICC profile for monochrome?! help me understand why that is even a thing.

Lot's of people reasonably expect monochrome images to look, well monochrome, without any tints or hints. An icc (colour)profile should get you to that standard, but a lot of canned profiles never quite hit the spot well enough. A custom profile should get that correct.

Remember that there's a whole fashion of making monochrome images with just a single element in colour, a red bus, yellow taxi etc, cliche maybe, but some like that sort of thing.

It's just a variation of optimising for what you print again.

[Alan Goldhammer]

In terms of a "monochrome" profile, I can only speak for Argyll which is the software I use for profiling.  It allows me to add B/W step patches to the profile set which will then be incorporated into the resulting ICC profile.  This will help smooth out the B/W gradient if the printer is non-linear.  Of course this uses the existing color printer and not a special B/W driver if the printer has that feature.

[Rhossydd]

It allows me to add B/W step patches to the profile set which will then be incorporated into the resulting ICC profile. 

Yup, the same process can be done with X-Rite software. Only you get a nice GUI to make life easier ;-)

[digitaldog]

Lot's of people reasonably expect monochrome images to look, well monochrome, without any tints or hints. An icc (colour)profile should get you to that standard, but a lot of canned profiles never quite hit the spot well enough. A custom profile should get that correct.

Remember that there's a whole fashion of making monochrome images with just a single element in colour, a red bus, yellow taxi etc, cliche maybe, but some like that sort of thing.

It's just a variation of optimising for what you print again.

That's not a monochrome profile, such profiles exist (Eric Chan created such profiles for Epson ABW).
You are referring to a profile with excellent gray balance and it's not a special use profile; all good ICC printer profiles should exhibit very good gray balance with fully colored images or partial colored images also containing neutrals:
http://www.digitaldog.net/files/2014PrinterTestFileFlat.tif.zip

[digitaldog]

In terms of a "monochrome" profile, I can only speak for Argyll which is the software I use for profiling.  It allows me to add B/W step patches to the profile set which will then be incorporated into the resulting ICC profile.  This will help smooth out the B/W gradient if the printer is non-linear.  Of course this uses the existing color printer and not a special B/W driver if the printer has that feature.

Not a unique feature. Many products allow including more neutral patches initially or post optimization to "target" gray balance. And then there are actually CMYK profiles generation (far more complex options) for affecting GCR and UCR.