Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: samueljohnchia on November 18, 2012, 02:39:19 am
-
These are my notes from a small experiment to investigate the properties of profiles that i1Profiler creates. It's going to be a bit long, as I continue to update this thread with further investigation.
Nobody tells us exactly what these sliders do. What exactly do they change? What are the downsides to these changes? Do we have to sacrifice some gamut for more smoothing, or maybe a higher dmax for a neutral grayscale? Under what circumstances are the changes useful? When are they irrelevant? Are there optimal values and suboptimal values for each slider? (That is, say never set the smoothness slider at XX value because the profile’s printing gamut suddenly becomes all wonky from a bug.) Do certain combinations of slider settings do things that are undesirable, hurt the profile quality, or maybe improve things beyond your wildest imagination?
I would like to clarify that I am not a color expert, or mathematician or scientist or the sort, so if anyone spots an issue that is technically flawed, I will be very happy to retract my statements, and I will re-investigate that aspect again.
Introduction
Andrew Rodney, in a review of the i1Profiler software on the luminous landscape website (http://www.luminous-landscape.com/reviews/accessories/x_rite_i1profiler_review.shtml), stated that “many of the provided sliders [in the "Profile Settings" module] produce tiny, almost insignificant differences from each end of the slider scale”. “Building profiles with both extremes for the Smoothness slider produced results that were less than a max deltaE of 1”. With all due respect to Andrew, I believe him.
I am curious if I can see a difference between various settings of the sliders when building RGB printer profiles. I made a series of profiles, from a measurement file I scanned from a target I printed, with varying inputs of the following sliders found in the “Profile Settings” module: Contrast, Saturation, Neutralize gray, and Smoothness. These sliders affect the rendering of the perceptual table of the printer profile. I am using i1Profiler version 1.3.2.
These settings remain constant for all the test profiles I generated. For "Tables", “Size A to B” is set to “Large”, “Size B to A” is set to “Large”, and “Granularity” is set to “16 bit”. Under "Advanced", “Chromatic Adaptation” is left at “Bradford”, a pointless setting since my illuminant used in the “Lighting” module is D50. ICC profile version is Version 2, and “Profile White point” is “Default”.
How I Made my Observations
Be aware that many of the following observations are SUBTLE. One method I use to better see the difference is to convert a granger rainbow with a sidebar neutral grayscale target (so I can see the effect on grays too) to various test profiles, layer the results in a non-color managed document, and use the difference blend mode and a strong curve adjustment to see where changes have occurred. An example of the target is attached.
I also use the ‘flash to compare’ method, where I create a custom proof setup for a test profile, and then change the proof settings to another test profile. Checking and un-checking the ‘preview’ checkbox ‘flashes’ the view. One method to judge the gamut of the profile would be to activate the gamut warning overlay together with a proof setup using the profile in question on a granger target in ProPhoto RGB color space. This is done in Photoshop.
I have looked at the Smoothness thus far, and will work on the others soon.
Observations from Studying the Test Profiles
Smoothness Slider
I created test profiles with the following inputs for the Smoothness slider, while Contrast, Saturation and Neutralize Grayscale were set to 0:
0, 10, 20, 30, 40, 50, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 90, 100.
When the “Smoothness” slider is set to zero, the resultant printer profile has the largest gamut. When it is set to 100, the gamut is much smaller. I do not know why the gamut has to shrink, but at this point, to my mind, one has to sacrifice gamut for smoothness. But it’s not so simple, more on that later.
Comparing 0 and 100, the profile with smoothing at 100 is visibly smoother mostly in the light yellow-green (I’m thinking sunlit grass, foilage) color areas. Reds are also visibly smoothed across most of the range. Oranges are affected to a lesser degree. Blues were somewhat smoother, mostly in the darker regions, but it's the most subtle change.
Looking at the gamut, in intervals of 10, it shrinks from 0 to 10, 10 to 20 and 20 to 30, but increases from 30 to 40 and 40 to 50 (but less so), then decreases from 50 to 60 (60 being smaller than 30), and continues to do so for 70, 80, 90 and 100. Still it is not so simple. Look at settings spaced 1 value apart.
It is interesting to note at this point that there are no changes to neutral gray tones until the smoothing slider hits 61. They remain unchanged at 0, 10, 20, 30, 40, 50 and 60. I didn’t try out settings in between those values, but I’m assuming that they do not change then too.
Smoothing Slider from 60 - 70
Ok, so the gray values are the same until 60. At 61, they lighten. It is mostly seen in the middle regions of the grayscale. From 62 to 63, they do NOT change at all, when compared to 61. Very interesting. Grays get darker going to 64, and lighter at 65. 64 compared to 60 is almost the same, with a 1 value change in one R, G or B channel only. From 66 to 69, gray values are almost the same as 65, with a 1 value change in one R, G or B channel only. Gray values lighten at 70. When I say 1 value change in one R, G or B channel only, I mean something like R130 G128 B126 changes to R130 G129 B126.
The profile gamut increased from 60 to 61, decreased going to 62 (62 almost identical to 60), increased at 63 (bigger than 61), decreased at 64, increased at 65 (bigger than 61 and 63), and decreased gradually from 66 to 67, 68, 69 and 70.
Smoothing Slider from 70 - 80
Gray tones get lighter from 70 to 71. From 71 to 79, increasing smoothing by a value of 1 resulted in minimal changes to grays. No more general lightening or darkening. Most grays do not change. Those that do exhibit a 1 value change in one R, G or B channel only.
Profile gamut increases from 70 to 71 (note that 71 gamut is greater than 61), decreased at 72 (72 slightly smaller than 70), decreased at 73 (by a very small amount), increased at 74 (very slightly too; 74 only slightly larger than 72), increased at 75, and more or less decreased (tiny differences) from 75 to 76, 77, 78 79 and 80.
Note that the gamut of the profile when Smoothness is set to 71 is larger than when it is at 61. So the assumption that gamut gets smaller when increasing the smoothing amount is not always true. Sometimes it increases. But 100 is much smaller than 0.
This is very very interesting behavior to me.
-
I'm interested in the effects of the sliders too.
To clarify what methods you're using;
How are you determining how gamut is changing ? As I understand it the tools in Photoshop aren't ideal for this and a profile inspection tool would be a more appropriate choice. It would also allow you to put some hard numbers to the investigation.
Later;
I've just spent a small time confirming your results with respect to gamut volume;
Setting the same as above using a 2033 patch set sample from an Epson 3800 on gloss media measured with an i1Pro.
Gamut volumes reported by Gamutvision
Smoothness slider setting > volume
0----888453
60---888453
61---806826
62---806801
63---888453
64---806351
65---806346
100--769910
These pretty much trend with your observations above.
Looking at the 3D gamut plots; It seems to be an overall shrinkage of gamut, with just the extreme very dark blue colour range being slightly more reduced than other areas.
Original reply edited to provide additional information.
-
Rhossydd, I assessed the gamut by looking at how many colors in a ProPhoto RGB granger rainbow target get clipped, indicated by the gamut warning overlay together with the profile loaded in view>proof setup>custom. This is done in Photoshop. I know it is not ideal. I do not have Gamutvision. I borrowed a friend's Colorthink Pro software and it thinks all the profiles have the same gamut volume, in the "Profile Inspector" module. Clearly they are not. Perhaps I don't know how to use Colorthink properly.
One thing about gamut volume numbers is that two profiles can have very similar gamut volumes, but may not include the same region of colors in gamut. Plotting them together in 3D will help us see where they overlap. That's where your software would be very revealing. Looking at the numbers alone won't tell the whole story.
I see that your profile at 0 and 60 are identical in volume. My profile created with smoothness at 60, clips a lot more colors than when at 0. Your results from 61 onward are similar to mine in that the gamut increases and decreases as the smoothing amount increases. It is not predictable.
It is very interesting to see it grow and shrink by a reasonable amount even with an increment of only one value. Looking at your gamut volume numbers, a simplistic conclusion could be if you have the smoothness slider at 65, setting it to 63 is a better idea. This is very very weird stuff to me.
-
We really want to view the gamut using ColorThink Pro. (http://www.colorwiki.com/wiki/Color_Management_Myths_26-28#Myth_26)
If memory serves, the settings only apply when using the Perceptual table. There have been a number of dot releases since my initial testing of some of the sliders.
The tests here seem to agree with my idea that a 0-100 scale (instead of 0-10) is X-rite just not thinking logical GUI design! The slider granularity currently used is just not necessary.
-
Hi Andrew, thank you for your input. I mentioned in the first post, I too think the sliders only affect the perceptual tables, not the colormetric or saturation tables. If someone knows otherwise, please correct me.
Graphing the profiles in 3D in Colorthink Pro agrees with my initial observations using gamut warning in Photoshop. It can be a little difficult when two profiles are barely larger than one another in gamut, to see where they differ. Would you like to have the profiles to take a look at them yourself? Not sure how the EULA applies here.
What I am thinking now is that the slider is not behaving predictably in small increments. Perhaps larger incremental amounts in its scale would be more logical. But given this kind of results, how would one choose an ideal setting for the smoothness slider? For example, gray values (equal R, G, B value combinations) do not get affected up to 60. At 61, they lighten visibly. Is this lightening better? More linear? More pleasing? I don't know. But hey, it darkens again at 64, and lighten again at 65. These are visible jumps in the gray tones, not the almost invisible jumps from 71 to 79. I am confused. Do I need more gamut or more smoothness? Sometimes I get both! Sometimes no. How do I determine the optimal slider position?
What happens with a different printer+paper combination?
-
At some point, you make a few profiles with the radically opposite settings, print a few images and then get the idea that these 'tools' provided are kind of silly as designed and delivered. That you roll one or two custom settings (as assets) and move on. A lot of the tools are marketing generated and of questionable usefulness. And yet do we have a tool here to edit the white point of a printer profile?
-
And yet do we have a tool here to edit the white point of a printer profile?
Latest version of i1Profiler lets you override the white point when generating a profile, if that's what you mean. Speaking of which, any idea why i1Profiler often gets the white point wrong for some papers? For instance I can measure a paper white of L* = 94 but i1Profiler will use L* = 92 by default. Does it have a good reason for doing this? Will it be to the benefit or detriment of my profile if I override and use the actual paper white value?
-
Andrew, what do you mean by "And yet do we have a tool here to edit the white point of a printer profile? "? Jeff pointed out that the latest version of i1Profiler allows a user to manually choose a white point. How is this useful for a critical user?
Jeff, thank you for pointing out this white point bug. I have yet to notice it, but I will pay close attention, since I am generating many many profiles, from various paper combinations. Just something weird I discovered early on, if I am generating profiles using the same measurement data, but with different "Profile Settings", I have to reload the measurement file data again in the "Measurement" module, otherwise i1Profiler will give me an error message saying that there is no available white point. This sounds stupid to me. How could it have 'lost' the white point?
-
I want to edit the white point after I build any profile (in some cases usually for CMYK cross simulations).
-
I want to edit the white point after I build any profile (in some cases usually for CMYK cross simulations).
Ah, thank you, Andrew. There is still a lot I don't know about, since I do not have experience with press printing, HP indigos, and digital lab photo finishing printers. I basically only have direct experience with pigment inkjet prints, so I never appreciated the need for that. But can't you take the profile made by i1Profiler and run it through PM5 to edit it?
I'm testing the contrast slider next, will update later.
-
Observations from Studying the Test Profiles
Contrast Slider
Introduction
I created test profiles with the following inputs for the Contrast slider, while Saturation and Neutralize Grayscale were set to 0, and Smoothness was left at the default of 50:
-50, -40, -30. -20, -10, 0 , 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50.
I chose to evaluate profiles created at those settings because the visible difference in results seemed most linear or obvious at those settings, except for when the slider was less than 0. I do not like the visual effect when softproofing profiles build with Contrast less than zero. The rendering of tones appear unnatural to me - the tonal range is flattened too much. Jumps of 10 in the negative direction produced visibly different results at every interval, if you are interested. I just didn't bother to look deeper. Too many things!
The exception is that the jump from 0 to 1 is relatively huge, and quite visible, up to a change of 4 values (getting darker) in all RGB channels. It is the single most significant change throughout the slider in the positive range. I find this quite unusual.
The change from 0 to 1 is about 400% larger than going from say 10 to 20, or 30 to 40. The difference of going from 0 to 1 is the same as going from 10 to 50. I do not suggest finessing the sliders between intervals of 10 when above 10 (I know, that sounds confusing). The differences are too small. However, finessing the sliders between 0 and 10 give about the same difference as say going from 10 to 20. So it looks like the granularity of the slider may be needed at near zero levels, almost like logarithmic behavior?
Observations Notes
Gamut
There is essentially no gamut change throughout the range of the slider.
From 0 - 1, there is a very very very slight loss of gamut in the dark greens. (less than the increase of in the light aquas)
From 7 to 8, there is a very very slight increase in gamut in the light aquas.
This additional clipping is so slight I cannot see it in 3D in colorthink. But the gamut warning in Photoshop indicates additional clipping. I am not familiar about the precision of the gamut warning in PS, so it may be wrong.
When I built profiles from a different printer+paper combination, and changing Contrast does not change the gamut at all.
Tone
The difference in tonal rendering from Contrast 0 to 1 is huge compared to going from 1 to 2, 3, 4, 5, 6, 7, 8, 9 and 10.
9 to 10 has a slightly less visible jump in contrast, where there is a less than 1 value increase in one or more R, G or B channels, in 8 bit terms. A 16bit/color channel target will indicate with greater precision where and what the differences are.
When I mention "RGB values", I was looking at gray values with equal amounts of R, G, and B. Bear in mind that other color combinations also experience change. I just looked at the grayscale for simplicity.
From 0 to one, RGB16 to RGB18 do not change in brightness. RGB values from 15 to 1 get darker. Values close to zero experience a 1 value change in one R, G or B channel only. Everything from 19 to 253 gets lighter. Values close to white experience a 1 value change in one R, G or B channel only. Note that the lightening and darkening here is very visible, not subtle. The maximum change is 4 values (darker) in all RGB channels.
When increasing Contrast slider amounts from 10 to 20, 20 to 30, 30 to 40 and 40 to 50, RGB values 236 to 255, 100 to 102, and 0 to 14 remain unchanged. From RGB103 to 235, grays lighten. from 15 to 99 they darken.
This makes sense as the RGB middle gray value of ProPhoto RGB based on the L* curve is about 100, 100, 100. It is interesting to note that the cut-off for the highlights is 20 values under 255, and shadows is 15 values above 0. I suspect greater highlight protection in the near white regions is employed because current inkjet printers show greater separation in highlight detail than in shadow detail. Flattening that extra information may not be desirable. This is my assumption only.
The maximum change is only 1 value in one or more R, G or B channels when going from 1>2>3>4>5>6>7>8>9>10>20>30>40>50.
I have not decided yet if leaving the contrast slider at 0 instead of at any other value is the most ideal setting. But I may just do so.
-
But can't you take the profile made by i1Profiler and run it through PM5 to edit it?
Not unless I boot into 10.6 or whatever was the last OS version that editor would run. Plus that editor really kind of sucks eggs IMHO. Lastly, where's the new editor or editing tools for the newer i1P package? Don't hold your breath. The pace of software development at this company is incredibly slow.
IF I were going to have to edit a profile, boot into an older OS to do so, I'd probably just go back to an older version of Photoshop and use the old and pretty awesome Kodak Custom Color profile editor (or boot OS9 or there about and have at it with ColorBlind Edit).
But the other point I'll make is that while it would be so great to have an effective, somewhat easy (OK just intuitive) profile editor, there are a few simple areas we just need some simple tools to take an existing profile and edit the back and white point of the profile.
-
The exception is that the jump from 0 to 1 is relatively huge, and quite visible, up to a change of 4 values (getting darker) in all RGB channels. It is the single most significant change throughout the slider in the positive range. I find this quite unusual.
I have not decided yet if leaving the contrast slider at 0 instead of at any other value is the most ideal setting. But I may just do so.
Thanks for taking what has to be a lot of time testing here. But I suspect what you might find is that there is really no logic behind the settings. At least what you report backs up what I suspected a very long time ago when I worked with the early product and came to the conclusion that the settings don't appear to be understood by those who coded them, let alone us users.
I use a setting of 40 contrast. Why? Well initially I set the Colorful preset/default if you will, that's were contrast stuck.
The other thing I'll suggest to any i1P user. IF a new version comes out, take some measurement data you saved and build the profile as you did in the past. Don't necessarily expect that drag and drop of an older profile will update all the workflows correctly. Test to see if anything new in the dot release caused a change compared to the older profile. Unless we are specifically told by X-rite that something in a new build affects the engine, don't assume it doesn't! Trust by verify.
-
Thanks for taking what has to be a lot of time testing here. But I suspect what you might find is that there is really no logic behind the settings. At least what you report backs up what I suspected a very long time ago when I worked with the early product and came to the conclusion that the settings don't appear to be understood by those who coded them, let alone us users.
Andrew, you are welcome. I'm just trying to gain some additional knowledge about what *I think* the sliders do. It's been rather quiet of late around the forums here. I remember some time back there were hot discussions about working spaces and all that. The fad seems to have died. I hope others will join in to help/verify what's going on their end. I would be grateful if Scott Martin, who is a great advocate of the quality of the i1Profiler profiling engine, could comment on his findings. I'm sure he'd had discovered this a long time ago. I also know that Tyler Boley of Custom-Digital (http://www.custom-digital.com/2011/07/summer-2011-update/) has also done some extreme testing and experimentation, even hacks, to achieve results that few of us ever can. If you are reading this, please share your findings.
I use a setting of 40 contrast. Why? Well initially I set the Colorful preset/default if you will, that's were contrast stuck.
I see less separation in the shadows with the contrast slider at 40 compared to 0. I may not use it at that value for my final profiles, but it is a subtle difference. Interestingly, above 0, the rest of the grayscale gets lightened, and seems to better match the tonality of the original synthetic grayscale. Of course your setup and requirements are different from mine and I am sure your profiles are excellent.
Whatever it is, it seems that setting the sliders at factors of 10 is fairly predictable. Incremental changes of 1 unit seem to produce highly unpredictable results, especially around the default value for that slider.
Default values for Contrast is 0, Saturation - 0, Neutralize Grayscale - 50 and Smoothness - 50.
Please, if anyone else out there also has i1Profiler, and can spare the time, it would be wonderful if you can verify these observations.
-
I have to admit I ran out of time and enthusiasm for this endless testing to find answers the people who coded the software should be able to provide. Also some support issues I worked on with them simply never got any attention. I'm very happy with the profiles, for the most part, the software creates. One question to complicate your tests...
Do these differences in settings results carry through when that same profile has been optimized? Optimization has proved absolutely necessary for it's CMYKOG profiles.
I'd love to dive back into this, and I'm happy to see your work, other things have taken priority lately.
Tyler
-
I'm hesitant to spend too much time going into all the specifics this week as I'm traveling but let me say that, as to the Perceptual rendering settings, I think it's important to separate color management from color correction, and therefore I try to find the settings that produce what I consider to be the best conversion that visually preserves the original image without "correcting" or "enhancing" it. With this in mind, I feel that the contrast setting is more of a subjective "enhancement" control as opposed to a valuable color space conversion adjustment. Make sense? There are surely uses for it but I doubt I'll ever use it. It's quite rare that it makes sense to combine color corrections with color space conversions.
The saturation slider is fantastic - it's one of the crown jewel technologies acquired from Monaco. The way it works is brilliant. Increasing saturation here doesn't effect or over-saturate what I like to call "mid-gamut" colors like skin tones. Instead, it controls the "edge gamut" colors. Think about super red saturated fall leaves and all the delicate veins and texture within the leaf. Use a low setting and the edge gamut colors are less saturated but full detail is present. Use a high setting and the edge gamut colors are much more saturated at the sacrifice of tonal detail. I think there's a sweet spot that gives the best of both worlds. But I won't spoil the movie - why don't you do your own testing and let's compare notes. :-] Same with Neutralize Gray.
Exciting to see you do this. Exhaustive testing like this is what separates the pros from the rest.
-
I have to admit I ran out of time and enthusiasm for this endless testing to find answers the people who coded the software should be able to provide. Also some support issues I worked on with them simply never got any attention. I'm very happy with the profiles, for the most part, the software creates. One question to complicate your tests...
Do these differences in settings results carry through when that same profile has been optimized? Optimization has proved absolutely necessary for it's CMYKOG profiles.
I'd love to dive back into this, and I'm happy to see your work, other things have taken priority lately.
Tyler
Tyler, I empathize with you. Nobody should have to test all the sliders at individual single unit increments. It is a pain in the ass to run through hundreds and thousands of profiles. The total number of combinations on the "Profile Settings" module is 14.4 billion, not including the variety of ways one can set the white point. If you include that, it is 9.2390544 Quintillion, or 9.2390544 billion billion. Seriously, that's not even including the optimization step.
I saw an early indication that using the sliders in combination can give unexpected results. This would be very painful to test extensively.
No, I haven't build CMYKOG profiles. You'll need a RIP to use that right? One cannot print or softproof in Photoshop with a 6 channel profile? I am on a Canon iPF8100 printer, so I don't think I can use a CMYKOG profile. For RGB printer profiles, Scott mentioned before that starting with a large enough patch set, one will not see any visible benefit from running the optimization step. I've been too preoccupied to test that.
By the way, have Scott or you noticed that when the patch sizes get really huge, like 5837, saving a mxf measurement file causes i1Profiler to crash? I have issues when I scan M0, M1 and M2 at the same time. It can handle M0 just fine, less total data. Like there is a memory limit on the mxf format. Smaller files have no issues at all.
-
I think it's important to separate color management from color correction, and therefore I try to find the settings that produce what I consider to be the best conversion that visually preserves the original image without "correcting" or "enhancing" it. With this in mind, I feel that the contrast setting is more of a subjective "enhancement" control as opposed to a valuable color space conversion adjustment. Make sense?
Scott, that is exactly my sentiment. If one wants more contrast for a particular image, the printing profile should not be the 'correction' method. Otherwise softer images that you want to print soft will also get hit with the contrast increase. I want the most 'neutral' setting for the tone curve, and with the best amount of separation, a long smooth tone in the grayscale, as visually linear as possible. I remember discussing with Joseph Holmes about the L* curve not being visually linear enough. Can't remember how he quantified it. I'll add contrast in post-processing if I need to. I tend to dislike "overly processed" imagery.
The saturation slider is fantastic - it's one of the crown jewel technologies acquired from Monaco. The way it works is brilliant. Increasing saturation here doesn't effect or over-saturate what I like to call "mid-gamut" colors like skin tones. Instead, it controls the "edge gamut" colors. Think about super red saturated fall leaves and all the delicate veins and texture within the leaf. Use a low setting and the edge gamut colors are less saturated but full detail is present. Use a high setting and the edge gamut colors are much more saturated at the sacrifice of tonal detail. I think there's a sweet spot that gives the best of both worlds. But I won't spoil the movie - why don't you do your own testing and let's compare notes. :-] Same with Neutralize Gray.
I'm on it. I realise that I left the neutralize gray at 0, which is not the default setting. I now wonder what that has done to my test results. Maybe I will not notice the 1 value changes in only one R, G or B channel anymore, I don't know. But then, 0 is no neutralizing, so I might see a greater effect if I left it at 50.
I worked late into last night creating my own custom targets to test the edge gamut of saturated colors. My initial reactions to the saturation are similar to yours. I will describe them in detail in a later post, and any other unusual findings.
Exciting to see you do this. Exhaustive testing like this is what separates the pros from the rest.
Scott, I'm not working at the moment so I can dedicate my time to doing this. I'd rather be doing something else. I was just at an art show, and there was this french photographer who made his own inkjet printing paper from Egyptian cotton, another who printed with the Piezography inkset, and still others on standard setups. They were selling their artwork at enormous prices. These are the "pros" too. But hey, they collect their money at the end of the day and go relax on their yacht or something. The print quality was absolutely rubbish. Worse still, I saw a lot of digital processing flaws, so even with a wonderful inkset like the piezography, I am sure I can do better with both eyes closed on my iPF8100+OEM inks. My eyes hurt so bad after the show. Honestly, I think nobody really cares that much to figure it out to the end. Perfection is just so elusive. It is so easy to give up and do something else. Seriously, there are times where I doubt I can see the differences if I didn't do a side-by-side comparison. Who is ever going to spot it at an artshow?? I'm not trying to show off here, but I think I am pretty good at visually judging prints objectively without comparisons, but this is just so subtle! I am reminded of a story I heard from an audiophile - "Can you hear the man cough at 2:05, fifth row, center column, third chair from the left?"
-
No, I haven't build CMYKOG profiles. You'll need a RIP to use that right? One cannot print or softproof in Photoshop with a 6 channel profile? I am on a Canon iPF8100 printer, so I don't think I can use a CMYKOG profile.
Yes, you need a RIP and yes Photoshop doesn't have much support for 6 channel profiles. You could use 6 channel profiles with your printer with the right RIP. The colorspace used is all about the RIP/driver - not the printer.
For RGB printer profiles, Scott mentioned before that starting with a large enough patch set, one will not see any visible benefit from running the optimization step. I've been too preoccupied to test that.
That's a big deal worthwhile test, IMO. There's a convenience/quality sweet spot for sure. And it comes way before 5000 patches. At some point having too many patches actually increases the chance that something will go wrong and bad measurements could be introduced...
By the way, have Scott or you noticed that when the patch sizes get really huge, like 5837, saving a mxf measurement file causes i1Profiler to crash?
Haven't seen that in a long time as that many patches is just insane with this engine. More patches doesn't equal better results like it used to.
-
Yes, you need a RIP and yes Photoshop doesn't have much support for 6 channel profiles. You could use 6 channel profiles with your printer with the right RIP. The colorspace used is all about the RIP/driver - not the printer.
Scott, will there be improved quality over what I can achieve with the 16 bit plug in from Canon? I read that Tyler has spent some time inventing a custom dither that is superior to the standard Epson driver. I find the dot formation by both Canon and Epson OEM drivers rather coarse. I can see the printed dots with my naked eye under reasonably bright light. I would be excited to know if I can tweak it away. There was also some talk about being able to control ink limiting to a very fine degree. Can I do it individually for all 12 ink channels? Would there be an advantage to the final profile quality? It would be wonderful to do that.
I'm also wondering how critical it would be to run the internal calibration for my Canon printer. If I can control ink limiting, would it be necessary to calibrate the printer if I am not running more than one machine?
Would running the calibration on the paper that I print on most often make the printer more linear for that combination? It would invalidate all my other current profiles though.
That's a big deal worthwhile test, IMO. There's a convenience/quality sweet spot for sure. And it comes way before 5000 patches. At some point having too many patches actually increases the chance that something will go wrong and bad measurements could be introduced...
I am also wondering if introducing a large number of grays and near grays actually improve the separation and neutrality in the grayscale. That would take me down the road of customizing targets. I haven't found any literature that describes in a scientific fashion an optimal way to select targets for printing photographic images. I am thinking that one could use around 2000 patches for color and near neutrals, and 1024 grays, 4 for each 8 bit integer, so the average would be more representative.
I'm wondering whether doing the gray optimization step (http://www.i1upgrades.com/2011/08/how-to-use-the-tc-2502-gray-optimization-chart/) after the profile has been built would be better/worse/similar compared to including all those grays with the initial target. I think Andrew advocates this quite a bit? I would really hate to run an optimization. It just adds more steps into an already complicated workflow, more paper, more ink, more drying time.
I also would like to test using 16bit targets but i1Profiler doesn't do those. Any workarounds available?
Haven't seen that in a long time as that many patches is just insane with this engine. More patches doesn't equal better results like it used to.
My initial testing some months ago show me that a 5837 profile has a visibly larger gamut, more neutral grayscale, higher dmax and "more accurate description of certain colors" (if my printer+paper combination cannot produce a perfectly smooth gradient for that color, the profile reflects that) than 2033 patch you recommended elsewhere on the forum. The downside is sacrificing some smoothness in the profile, which appear to be correctable by the smoothness slider. But I haven't tested this extensively with different printer+paper+driver combinations to know for sure.
-
Scott, will there be improved quality over what I can achieve with the 16 bit plug in from Canon?
I don't think so. The plug-in is fantastic. Ever noticed that the soft proofing is better than Photoshop's?
There was also some talk about being able to control ink limiting to a very fine degree. Can I do it individually for all 12 ink channels? Would there be an advantage to the final profile quality? It would be wonderful to do that.
Sounds like StudioPrint is the RIP for you. Have fun geeking out but don't expect to get significantly better quality or anything.
I'm also wondering how critical it would be to run the internal calibration for my Canon printer. If I can control ink limiting, would it be necessary to calibrate the printer if I am not running more than one machine?
Calibration is too important to ignore, IMO. It will allow you to achieve incredible consistency over time and across head replacements.
http://canonipf.wikispaces.com/message/view/FAQ/12791457
I am also wondering if introducing a large number of grays and near grays actually improve the separation and neutrality in the grayscale.
Yes, but you don't really need that many more grays... Use i1P's patch generator and find a set that includes a ton of grays. If you increase the number of patches by one you'll see that the grays increase up to apoint where it falls back down to a minimum value. Try it - I think you'll find the intervals where there are a whole bunch of grays (like 2033).
That would take me down the road of customizing targets.
Yeah, and that's just an academic exercise, that's not likely to produce significantly better results, IMO. I have spent stupid amounts of time to make my own 16 bit i1P targets with lots of grays and a visual layout like Bill Atkinson's targets used to have. You can see the small one's at http://www.on-sight.com/downloads/ But the differences are small in comparison to what you can get by using i1P's targets at those intervals where you find a bunch of grays. I'd stick with that.
I also would like to test using 16bit targets but i1Profiler doesn't do those. Any workarounds available?
Yes but there's not much reason to do it. I actually have my own 16 bit targets and 16 bit references files that I made with the help of XRite developers. I use ColorPort to measure them and i1P to process them (which it does do in 16 bits). But there's not much difference doing it this way. If anything it allows you to get a little better quality out of fewer patches. Going forward ColorPort is hard to rely upon.
-
I'm wondering whether doing the gray optimization step after the profile has been built would be better/worse/similar compared to including all those grays with the initial target. I think Andrew advocates this quite a bit? I would really hate to run an optimization. It just adds more steps into an already complicated workflow, more paper, more ink, more drying time.
I did this (optimization) for one of my profiles, and there was significant benefit. The profile in question was originally created with a roughly 3400-patch target which had a lot of gray or near-gray patches; but when printing a test image that had a monochrome circle gradient, there was some visible banding (this particular test image is extremely unforgiving). After using the optimization step, the improvement was obvious, with the circle gradient being smooth.
There's a problem though. The optimization workflow is extremely buggy. For one thing the loading/saving of patch-sets/test-charts doesn't work well. But an even bigger problem is that the optimization workflow doesn't bother to arrange the target charts so that adjacent patches have enough difference to register as distinct patches when scanning a row at a time. The only way I've found around this is to use the "legacy charts" option with my i1Pro2, so that there are hard borders between the patches. This works fine, except for the fact that you can only scan legacy charts in M0 mode. So for profiles that you created using dual-scan for OBC, it won't work. So for the few papers I have that use the OBC workflow I can't optimize the profiles. Fortunately those particular profiles are smooth enough that it's not really an issue.
Overall I'd have to say X-Rite should really be ashamed of themselves for just how buggy this software is. Sure the profiling technology is pretty good, but the actual Windows software is just poorly written. I can't tell you how many times I've gotten errors or outright crashes when generating OBC charts, profiles, etc. If you restart the app and try again it will usually work, but that's assuming you saved what you needed to be able to resume after restarting. The program must have memory leaks out the wazoo, not to mention bugs that cause the state of your workflow to get corrupt in various scenarios. Going back/forth through the workflow steps is particularly problematic, I've also gotten that "cannot generate white point" error message many times before.
-
The plug-in is fantastic. Ever noticed that the soft proofing is better than Photoshop's?
Do you really find the plugin's soft-proofing useful? If there were some way to get it to display full screen that would be one thing, but that tiny image preview is just too small on my 24" display. I know you can zoom in, but that just seems like an extremely tedious way to do it.
-
I don't think so. The plug-in is fantastic. Ever noticed that the soft proofing is better than Photoshop's?
That's good to know. I had the feeling that the plug-in's softproofing seems a tad more accurate, but it doesn't offer a simulate paper white preview. I also thought that it doesn't lift the blacks as severely as Photoshop's softproofing, with simulate black ink on.
Jeff, I have no problems expanding the plug-in window to get a very large preview of the soft-proof. Well, it's quite big to me on my 27 inch display, no smaller than fitting the image to display area in Photoshop. Yeah, zooming in is annoying. I didn't even know I could do that until you pointed it out!
Calibration is too important to ignore, IMO. It will allow you to achieve incredible consistency over time and across head replacements.
Oh no, all the re-profiling to do!! Argh...
But the differences are small in comparison to what you can get by using i1P's targets at those intervals where you find a bunch of grays. I'd stick with that.
Do the custom targets produce better, smoother or more neutral profiles?
-
Jeff, I have no problems expanding the plug-in window to get a very large preview of the soft-proof. Well, it's quite big to me on my 27 inch display, no smaller than fitting the image to display area in Photoshop. Yeah, zooming in is annoying. I didn't even know I could do that until you pointed it out!
The version of the 8300 plugin I'm running on Windows has a fixed-dialog border and cannot be resized. However after reading your reply I went back to double-check this, and noticed that there _is_ a maximize button on the dialog. Never noticed that before (although I have tried resizing the window before). So while I can't resize, at least I can maximize the window. The only downside is the combo-box for selecting ICC profile doesn't expand when the dialog is maximized which means I can tell some of my profiles apart due to the way they're named.
-
The version of the 8300 plugin I'm running on Windows has a fixed-dialog border and cannot be resized. However after reading your reply I went back to double-check this, and noticed that there _is_ a maximize button on the dialog. Never noticed that before (although I have tried resizing the window before). So while I can't resize, at least I can maximize the window. The only downside is the combo-box for selecting ICC profile doesn't expand when the dialog is maximized which means I can tell some of my profiles apart due to the way they're named.
Jeff, that is exactly how it works on my end too.
-
Those who print in color, I hope you dig this:
Observations from Studying the Test Profiles
Saturation Slider
Introduction
I created test profiles with the following inputs for the Contrast slider, while Saturation and Neutralize Grayscale were set to 0, and Smoothness was left at the default of 50:
-50, -40, -30. -20, -10, -9, -8, -7, -6, -5, -4, -3, -2, -1, 0 , 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50.
I also created my own custom targets to study how well colors separate at near clipping. The target I created has a background of a color primary at clipping, that is 255 in 8 bit integer form. There are 20 small squares on the target. They run 254, 253, 252, 251 250, 249 etc. from left to right, top to bottom. The 20th square is 235. An example of the target is attached. I created separate targets for Black, White, Red, Green, Blue, Cyan, Magenta and Yellow.
Observations Notes
Gamut:
There is essentially no change to the gamut of the profile at all settings on the Saturation slider.
Tone:
Moving the Saturation slider from 0 to 1, there is a relatively significant increase in brightness in the grayscale. Everything from and including 0 - 224 brightens.
The most significant change is at RGB59. An increase of 5 values in all R,G,B channels. This is also true when moving from 0 to negative 1.
The grayscale brightness does not change much from Saturation 1 - 50. At intervals of 10, a maximum difference of 2 RGB values was noted in one or more R, G, and B channels.
However, the overall lightening jump from 0 - 1 remains. This is also true when moving from -1 to -50.
Separation of Colors Near Clipping:
Reds
- S40 gives the greatest separation of the positive scale, down to 250,0,0. Wierdly, lower or higher values than 40 separate less well. Reds get darker as saturation amounts increase.
- S -40 gives the greatest separation of the entire scale, down to 254, 0, 0. Reds lighten when saturation decreases.
Greens
- S20 gives the greatest separation, down to 0, 251, 0. Greens turn darker and cooler when saturation amounts increase.
- Negative saturation settings do not improve separation, and quite often makes it worse.
Blue
- S20 gives the greatest separation, down to 0, 0, 254. Blues also lighten significantly when saturation amounts increase.
- negative saturation settings do not improve separation, and always makes it worse.
Magentas
- I can saturate up to S40 without loosing separation, down to 253, 0, 253. Magentas get darker as saturation amounts increase.
- Magentas separate similarly well at negative saturation settings. Magentas get lighter as saturation amounts decrease.
Yellows
- I can saturate up to S50 without loosing separation, down to 253, 253, 0. Yellows get darker as saturation amounts increase.
- S -20 gives the best separation of the entire scale, down to 254, 254, 0. Yellows lighten as saturation amounts decrease. Max yellow (255, 255, 0) becomes unnaturally light for negative saturation then. This effect is very very ugly.
Cyans
- I can saturate up to S20 without loosing separation, down to 0, 251, 251. Cyans get darker as saturation amounts increase.
- S negative 30 gives the the greatest separation of the entire scale, down to 0, 253, 253.
- Max Cyan becomes unnaturally light for negative saturation. This effect is ugly, but not as bad as yellow.
Preliminary Conclusions
Choosing the optimal Saturation slider amount requires one to know what colors are the most important to one. Settings that are optimal for one color primary can produce results that are quite horrible for another. Generally speaking, I do not like the visual effect of negative saturation amounts, especially for the yellows. One can still retain very good separation at positive amounts of the saturation slider, for all the color primaries listed above. Red may be an exception.
-
Observations from Studying the Test Profiles
Neutralize Grayscale Slider
Introduction
I created test profiles with the following inputs for the Neutralize Grayscale slider, while Contrast and Saturation were set to 0, and Smoothness was left at the default of 50:
0 , 1, 2, 3, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100
Observations Notes
Gamut:
There is essentially no change to the gamut of the profile at all settings on the Neutralize Grayscale slider.
Tone:
Moving the Neutralize Grayscale slider from 0 to 1, there is a relatively significant increase in brightness in the grayscale. Everything from and INCLUDING 0 - 224 brightens.
The most significant change is at RGB59. An increase of 5 values in all R,G,B channels. This is also true when moving from 0 to negative 1.
The grayscale brightness does not change much from Neutralize Grayscale 1 - 100. Often, no change is noted.
Color Changes:
Comparing 1 and 100, a difference of 1 RGB value was noted in the blue channel, for every value above 93, just below middle gray (100) for ProPhoto RGB, based on the L* curve.
Comparing 10 and 100, a difference of 1 RGB value was noted in the blue channel, for every value above 100, exactly at middle gray for ProPhoto RGB, based on the L* curve.
Preliminary Conclusions
The paper that I printed on is Harman by Hahnemuhle Gloss Baryta, which has moderate amounts of OBAs in the paper base. Ethan has noted that the slider has more effect on papers containing lots of OBAs or when using "older tungsten-illuminated instruments (i1Pro, iCColor, Spectrolino) (http://www.luminous-landscape.com/forum/index.php?topic=62905.msg506857#msg506857)". I do not use papers with very high OBA content, so I cannot comment on that situation. The discussion in that thread does make logical sense though.
I must move the Neutralize Grayscale slider from 1 all the way to 80 before I can detect a difference in 8 bit integer precision. In 16 bit precision, there is a small difference at every increment in the slider, but no one should be able to notice the increment visually. The single greatest difference is going from 0 to 1, a huge brightness increase.
The Neutralize Grayscale slider also affects colors, with up to 4 values in difference in the blue channel, less in the red and green. Increasing blue is logical, to neutralise the otherwise too-warm interpretation of the profile, when a UV-cut measurement is not made. The colors that are affected the most are the dark yellows and greens.
-
I wonder how the Neutralize Gray slider interacts with the OBC module, but I suspect figuring that out might be difficult.
-
I wonder how the Neutralize Gray slider interacts with the OBC module, but I suspect figuring that out might be difficult.
No Jeff, it is not difficult. I just ran some more profiles to investigate that. I'll appreciate it too if you could run these tests on your favourite printer+paper combination, and let us know if you observe something similar.
Regarding OBC profiling, the Neutralize Grayscale behaves similarly, except that the big brightness jump from 0 to 1 is not present. I still get a change of up to 2 values in one or more R, G, and B channels. Don't know why.
I'm guessing that if one picks patches from the OBC test chart that result in non-neutral grays when building the profile, the neutralize gray slider will still have visible effect. I'm not sure if i1Profiler is looking at the AtoB or BtoA perceptual table (possibly both) to figure if a gray is neutral or not. It may be looking at lab values for all I can tell. If grays are already very neutral to begin with, then it has much less of an effect. That's all I can tell now.
-
On the Epson printers I have (3880, 4900), using a custom 1728 patch target from i1P, I do see a slight benefit to the 2500 patch gray optimization. It only kicks in with Perceptual and it's subtle. But if you have an iSis like I do, it's worth the time (media, that's your call). And so far, no customers who've had optimization have indicated anything worse than the optimized profile seemed to output like the original. No harm done.
You have to proceed in the proper order here if you also want to use OBC module. You'd do OBC then Optimization. Can't Optimize then do OBC (why I don't know). OBC is pretty slick and useful. But you tie the profile to one viewing condition. The visual based system works pretty well (we need more of this. I'd love to see the Macbeth used more in this product, the current implementation is really half baked).
All these sliders for profile creation and at least for me, when soft proofing, I pick RelCol 9 times out of 10.
-
I wonder how the Neutralize Gray slider interacts with the OBC module, but I suspect figuring that out might be difficult.
If this is supposed to be something like we had in ProfileMaker Pro, it's supposed to account for very non white (neutral) papers. Like newspaper stock. I'd suspect OBC would be another process that may or may not be necessary,
-
All these sliders for profile creation and at least for me, when soft proofing, I pick RelCol 9 times out of 10.
For fiber-gloss papers, so do I. So far I haven't found any combination of settings for which the perceptual rendering doesn't change the contrast of the image, which I don't like. With the Contrast slider set anywhere from 0 or higher, the bottom half of the tonal range gets at least slightly darkened, while negative Contrast slider values avoid this but have the effect of flattening the upper tones. It's really strange because in PMP/i1Match the perceptual rendering always lightened the image.
Matte papers are different though. Rel-col will almost always cause too much detail loss in the darker areas. And given the tonal compression needed to fit the more limited gamut/dmax of matte papers, the perceptual rendering doesn't increase contrast and the lower tones don't get darkened. I almost always end up preferring perceptual intent for matte papers.
-
...
Exciting to see you do this. Exhaustive testing like this is what separates the pros from the rest.
well this comment immediately followed mine, where I stated my patience for flushing out every interface aspect and troubleshooting problems had waned. Exhaustive testing is done here all the time, and was and is still done for this software. But it is prioritized to the issues most likely to the most significantly effect results in the prints. I confess I did not do all 14.4 billion setting combination tests, perhaps the real pros here did.
I would not advise others to go down the n channel, RIP, path as I have, you will encounter even more problems with the software to flush out on your own, not to mention entirely new issues leapfrogged by the RGB workflow. I'd only advise it if you are ready for a significant learning curve to take over your life for a while, little good advice from anyone else, and have very real printing challenges that can be addressed only by diverting from the RGB workflow. This includes many things we take for granted, like the tools we all use for gamut volume info, etc.. new questions even there in n channel...
Again, I appreciate your testing, and sharing your results.
Tyler
-
On the Epson printers I have (3880, 4900), using a custom 1728 patch target from i1P, I do see a slight benefit to the 2500 patch gray optimization. It only kicks in with Perceptual and it's subtle. But if you have an iSis like I do, it's worth the time (media, that's your call). And so far, no customers who've had optimization have indicated anything worse than the optimized profile seemed to output like the original. No harm done.
You have to proceed in the proper order here if you also want to use OBC module. You'd do OBC then Optimization. Can't Optimize then do OBC (why I don't know). OBC is pretty slick and useful. But you tie the profile to one viewing condition. The visual based system works pretty well (we need more of this. I'd love to see the Macbeth used more in this product, the current implementation is really half baked).
All these sliders for profile creation and at least for me, when soft proofing, I pick RelCol 9 times out of 10.
Andrew, I do not have an ISis. I use an i1 Pro 2, so scanning increasing amounts of patches really add up. I don't do remote profiling or many printer+paper combinations, so I don't really need any automation yet. I'll test the Optimization process in the near future.
I also tend to find that an OBC-ed profile tends to end up with the grayscale looking very blue. The more OBAs present, the bluer the grayscale appears in the soft proof. It isn't really what prints out though. About tying it to one viewing condition - lots of prints have to be viewed in changing/different lighting conditions. Do you find OBC useful in those situations?
-
The more OBAs present, the bluer the grayscale appears in the soft proof. It isn't really what prints out though.
If the soft proof doesn't sync up after OBC, then it sounds like a bug or issue with the preview portion of the profile. Do you see any visual difference in a soft proof from identical profiles where the ONLY difference was OBC correction?
-
If the soft proof doesn't sync up after OBC, then it sounds like a bug or issue with the preview portion of the profile. Do you see any visual difference in a soft proof from identical profiles where the ONLY difference was OBC correction?
My words may have sounded too harsh. The differences are subtle.
There ARE visual differences in a soft proof from identical profiles were the only difference was OBC correction. Obviously, this is to correct for the overly yellow profile, which was compensating for the too-blue-to-the-software+spectro effect. I want to see differences, otherwise I would think OBC is not working.
The blue-shift difference is less in the print I suspect because of visual constancy. My eyes are compensating to neutralize what I *think* are neutrals.
The relative difference is bit like seeing the blacks go all gray and flat when softproofing a matte print with black ink simulation- none of the matte prints I make ever print out that flat in real life. Sometimes, they can even seem denser than glossy blacks, because of surface reflections.
-
Maybe the A to B table(s) need some work after OBC? Not that we have a tool for that in the package.
-
In response to the comments above about improvements after optimizing profiles in i1Profiler, I attempted my own investigations.
Introduction
There has been some discussions going on the luminous-landscape and other forums about optimizing profiles in i1Profiler. The general view is that if one starts out with a large number of patches for building the initial profile, say around 2000 patches, then the optimization routine does not show much fruit. This is true for when the optimizing patches are generated automatically by the “Smart patch generator”. Andrew Rodney noted that the maximum differences were less than 1 deltaE. (http://www.luminous-landscape.com/forum/index.php?topic=65827.msg521383#msg521383) That is hardly any difference. However, Andrew and others have seen a marked increase in gray neutrality and smoothness when using the 2502 spot grays optimization file and process detailed here: http://www.i1upgrades.com/2011/08/how-to-use-the-tc-2502-gray-optimization-chart/ (http://www.i1upgrades.com/2011/08/how-to-use-the-tc-2502-gray-optimization-chart/).
I want to test whether optimizing a profile with patches generated by the smart patch generator result in any improvement. I also want to test whether optimizing with the custom 2502 spot grays will result in any improvement. I tested this on two profiles, which I will call "Profile A" and "Profile B" for simplicity.
Both profiles were built from targets containing over 3000 patches, but less than 4000 patches. Both targets contained a significant number of neutrals, more than 329 neutrals and near-neutrals, which is the maximum number that i1Profiler can generate automatically.
Test Results
Optimizing Profile A with 2502 grays
Grays
• After optimization, grays became noticably cooler, especially in the highlights. Smoothness was slightly but visibly improved.
• There was a maximum increase of 5 values in the blue channel (the maximum increase occurred from RGB114 to 122), and a max value of only 4 in the R or G channels. This explains the cooler grays.
• There was also slightly improved separation in the dark grays, close to black, in the RGB 0 to 60 range)
Colors
• After optimization, there was a noticeable but slight increase in separation in the blues near clipping (Blues 235 to 255).
• After optimization, there was huge separation loss in the yellows near clipping (Yellows 240 to 250)
Optimizing the Profile B with 1260 patches generated by the “Smart patch generator”
I generated a 1260 patch chart using the “Smart patch generator”, which happen to contain a number of colors and a fair proportion of neutrals and near-neutrals, in a ratio of about 2 part colors, 1 part neutrals.
The optimized profile has a gamut volume of 766 217, calculated by ColorThink Pro.
The original profile has a gamut volume of 768 465, calculated by ColorThink Pro.
Grays
• Gray values hardly changed. A maximum increase of 2 values in the blue channel and a maximum change of only 1 value in the R or G channels was noted.
• Grays did not seem to get lighter or darker.
• Grayscale did not become smoother.
Colors
• Magentas were the most severely affected. It appears that i1Profiler attempted to correct the magentas to give a smoother result, but a large portion of it was lightened, and a small but visible transition to the darker shades appeared. This is arguably a worse result.
• Dark Oranges and greens appear to be slightly smoother, but the difference is so slight I can barely make it out in the softproof. Only by converting a granger rainbow to the original and optimized profiles and comparing them as two layers in a non-color managed Photoshop document using the difference blend mode, did I first notice this difference.
• There was huge separation loss of the yellows near clipping (Yellows 235 to 255) after optimization.
• Blue separation was very good to begin with, so optimization did nothing to improve it.
Optimizing Profile B with 2502 grays
Optimization failed. X-rite said that there was an error creating profile, despite me making 4 different measurement scans, and multiples tries rebooting the computer, and re-launching i1Profiler. I suspect that the memory limit on i1Profiler was reached, because of too many patches in total (3256+2502 patches). I get the same error after I have asked i1Profiler to generate about 8 large-patch profiles, and I have to re-launch the application to clear the memory.
Preliminary Conclusions
The Profile B did contain a fair bit more spot gray patches than Profile A. This resulted in a profile, which before optimization, that is already far smoother than any of the i1Profiler-automatically-generated-patches-based profiles I have created, and is even smoother than Profile A after optimization with 2502 grays. This is outstanding performance.
It seems, to my way of thinking, that selecting optimal patches for profiling does result in a superior final profile, and having a superior profile is better than trying to optimize a sub-optimal profile. More on optimal printer targets later.
I wish to make it clear that Profile A already has a pretty smooth grayscale gradient with almost no banding to begin with, and it is as smooth as the i1Profiler 5837 patch profile. The additional smoothness I see from either the optimization process or the superior custom target is subtle, but visible.
This investigation also informs me that the optimization that i1Profiler does is detrimental to the separation of high saturation yellows, a color which I find for my Canon iPF8100, is very hard to separate to begin with. I am also not happy with the way it dealt with the magentas.
Getting a smoother grayscale with Profile B took far less time, ink and paper, and produced a smoother grayscale than Profile A after optimization.
All in all, I have to say that an optimized version of an already very good printer profile, is worse than the original profile.
An optimized version of a suboptimal profile probably will result in an improvement, but I do not think creating a suboptimal profile to begin with is a good idea. I also question whether the gains of optimization, at the cost of additional ink, paper and time, is worth it.
-
All these sliders for profile creation and at least for me, when soft proofing, I pick RelCol 9 times out of 10.
There has been some research (http://cias.rit.edu/~gravure/tt/pdf/cm/TT5_Jorge01.pdf) that when using the relative colormetric intent, with Adobe's black point compensation turned on, colors are shifted in hue and saturation more than they would, compared to the perceptual rendering intent. Other research (http://www.gamutvision.com/docs/blackpoint.html) that I have read, including Adobe's technical article on the implementation of black point compensation, also indicate that even the lighter colors (up to density 0.5 or so) are also affected.
Basically, black point compensation turned on results in less accurate colors in terms of hue and saturation.
So far I haven't found any combination of settings for which the perceptual rendering doesn't change the contrast of the image, which I don't like. With the Contrast slider set anywhere from 0 or higher, the bottom half of the tonal range gets at least slightly darkened, while negative Contrast slider values avoid this but have the effect of flattening the upper tones. It's really strange because in PMP/i1Match the perceptual rendering always lightened the image.
I haven't noticed this loss of separation in the dark values with the perceptual rendering intent, when using a well-made profile. Actually I have found that rel col with BPC results in a grayscale that is less smooth, and less like the original tonality of the synthetic target. With the contrast slider of i1Profiler set to 0, There is no darkening of the low values, as I have discovered with additional testing. This was not mentioned in my previous post about the Contrast slider. Moreover, the printing gamut of perceptual is larger than rel col for i1Profiler's profiles.
Scott has been saying for some time now that Monaco Profiler, and now i1Profiler, is to be noted for simply outstanding perceptual output (http://www.luminous-landscape.com/forum/index.php?topic=43876.msg368218#msg368218). But lots of folks think that perceptual compresses in gamut colors too much, and so resort to printing with rel col. If you are printing scientific or technical images with fairly light colors and no shadows of any sort, I wouldn't hesitate to print with relative colormetic, without black point compensation on. If you are printing images with highly saturated color, with important shadow information, I think perceptual may be a better option.
Scott, I would greatly appreciate your input and advice on the advantages of printing with perceptual, when one has a good quality printer profile.
-
There has been some research (http://cias.rit.edu/~gravure/tt/pdf/cm/TT5_Jorge01.pdf) that when using the relative colormetric intent, with Adobe's black point compensation turned on, colors are shifted in hue and saturation more than they would, compared to the perceptual rendering intent.
I don't really care. When I say I prefer RelCol 9 times out of ten, this is soft proofing images and doing so with lots of profiles even before Photoshop could even soft proof. IOW, I pick this rendering intent a lot time with lots of actual images and lots of differing profiles (RGB, CMYK and for lots of output devices).
What's important to understand is that profiles are kind of dumb. They only see images as huge piles of solid colors. They don't understand colors in context and treat a super colorful scene just like a black dog on a pile of coal. I could care less about dE values here, I care about how my images look in context when I toggle not only RelCol vs. Perceptual but Saturation too (it can work quite well).
My take is to always leave BPC on. If off, and the profile needs it, it helps tremulously and when the profile doesn't need BPC, it plays no significant role PLUS when BPC is on, I prefer the rendering.
Here's what an ICC profile "see's" if you well. One is RelCol, one is Perceptual. I'm sure one has a closer dE value to something I don't care about in this discussion. Which is better? Until you zoom out and view the entire image and then toggle the intents to select one, it's not too easy to pick a rendering intent correctly for your needs.
(http://digitaldog.net/files/WhatProfilesSee.jpg)
Having dE values IS useful, but not as useful as just printing a lot of images with the profile after at least picking a preferred rendering of your image. And if you're going to run images through profiles to evaluate the color quality, the test images are kind of important. If you only want to deal with 16 or less, you need something like the Roman 16's which pile a lot of really useful visual image info into a small group of high quality test images. Print two sets with both rendering intents, or just LOOK at a good soft proof. You'll get a lot of good information this way that no dE report can provide.
-
The optimization process came from ColorMunki technology where one would build a profile with a tiny set of patches (50 initially). I'm stunned at this technology. For Munki profiles it works really well! I recently built a profile for a friend using all the usual i1P tactics: 1789 custom patches, an iSis in dual scan mode, the 2500 patch optimization. This fellow built a profile for the same printer using ColorMunki and 100 patches. Mine was better. But the differences were pretty small and while the dE values might be interesting to us geeks, when the various prints made using both profiles were examined, the differences were small enough that I suspect a pretty significant segment of people viewing the two wouldn't notice the differences.
Optimization on the i1P profile above made tiny but visual (if you looked) differences. As I've said in the past, it's great if you have an iSis or you have a lot of time to scan targets and you are super picky and have a good eye. Otherwise I'll offer it but tell people not to expect big differences.
I'd rather NOT have to resort to this process. I'd rather just measure all the patches, build a profile and be done. I suspect that if a subset of the 2500 gray's were incorporated into the original target, this secondary optimization, even for the ultra picky might be moot. I'm very impressed X-rite has this technology. It wasn't aimed for and doesn't really help the advanced i1P user anywhere like it does the ColorMunki user (or someone with i1P who would do something like build a profile with an initially small set of patches).
As for a Perceptual intent, yes, i1P's are better then the legacy app's. This illustrates what was talked about earlier where such differences in all profile making software product in terms of "do anything you want that you think looks good" with a perceptual intent. Think of it a lot like film stocks from various companies. Is Velvia better, more accurate than Kodachrome or Agfachrome (OK that was an ugly film stock <g>)? They all had their own unique 'looks' which is similar to what we see with Perceptual rendering intents.
-
I don't really care. When I say I prefer RelCol 9 times out of ten, this is soft proofing images and doing so with lots of profiles even before Photoshop could even soft proof. IOW, I pick this rendering intent a lot time with lots of actual images and lots of differing profiles (RGB, CMYK and for lots of output devices).
...
Andrew, thank you for your valuable advise. I fully agree that testing on good quality and useful targets as well as real world images would tell a more realistic story than simply looking at the numbers alone.
I also think that BPC is useful, and it renders photographic imagery in a more pleasing manner in the deep shadows.
In my own testing, I noticed that relative colormetric introduces banding into the yellows where perceptual does not. This is irregardless of having BPC on or off. This is only for profiles made by i1Profiler. Have you seen this issue?
-
Ah, I figured it out. I don't think anyone else should have this problem. Please ignore my question.
-
This is only for profiles made by i1Profiler. Have you seen this issue?
Only on synthetically created images like a Granger Rainbow. There's areas of banding in many area's, I don't expect to see perfect smoothness.