Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: alan9940 on August 11, 2018, 12:16:21 pm
-
Hello All,
Just read with great interest Mark Segal's recent article on the dark art of scanner profiling. Since I have a license for Ai Studio 8, the auto-calibration tool is quite attractive to me. I currently have a few targets, but the manual process of doing the profiling and ensuring everything was setup correctly in the scanner software was always a bit daunting to me. Anyway, does anyone know what the difference is between a "Calibration Target" and an "Individually measured Target", as sold by Silverfast?
Thanks!
-
Hello All,
Just read with great interest Mark Segal's recent article on the dark art of scanner profiling. Since I have a license for Ai Studio 8, the auto-calibration tool is quite attractive to me. I currently have a few targets, but the manual process of doing the profiling and ensuring everything was setup correctly in the scanner software was always a bit daunting to me. Anyway, does anyone know what the difference is between a "Calibration Target" and an "Individually measured Target", as sold by Silverfast?
Thanks!
Hello Alan,
To use the SilverFast Auto IT8 process you need to use one of their targets because they are barcoded and the application needs that barcode to create the profile.
When they print the targets, they do so in batches and there could be minor variations in the colour values of the patches from one target to the next. The "calibration target" uses a reference file containing representative colour values for the whole batch measured from one (or perhaps several averaged) of the prints, whereas the "hand measured" target has a reference file containing the colour values specific to that particular copy of the target because it is individually measured. In all other respects, they are the same target. While a hand-measured target having its own specific reference file is ideal, from what I've seen in the past, I'm not convinced it's worth the extra cost.
-
Alan, +1 on Mark's paper.
LSI has stated that they make the Standard Targets in a "very controlled" process of 50 at a time. This dates back some years. They hand measure one of them, so I think but Mark could be right about measuring more than one and averaging them. They apply the reference file to the batch.
Back when I bought an Ektachrome and Kodachrome [lucky me] 35mm slide targets they were $50 US for the batch targets. If memory serves me I would get a dE of 0.7 or 0.8 on the Ektachorme batch target. Later I got a measured Ektachrome target in 7-2013 for $216 US. I would get dE of 0.4 and occasionally a 0.5.
This is well below what some say 4.0 dE is acceptable. Can't see the difference.
Interestingly Mark has stated that the canned profiles are also very good, at least for the Epson V850.
I'm wondering wether my measured Standard Target is better than a batch Advance Target. As I don't want to spend the extra cost again, and I'm not sure they offer the Measured Advance Target.
-
It would be unusual to see a systemic difference of results between a profile with their stated dE 0.4 versus one with their stated dE 0.8.
I'm not sure whether they average several targets for the batch measured ones - perhaps yes, perhaps no. I just suggested it as a possible alternative to using one for representing the whole batch.
They do offer individually measured Advanced targets, but they are substantially more expensive than the batch versions.
I haven't analyzed an individually measured Standard versus a batch Advanced, but my hunch based on the evidence in the article is that the batch Advanced should still be inherently better than either Standard in several respects.
-
I haven't analyzed an individually measured Standard versus a batch Advanced, but my hunch based on the evidence in the article is that the batch Advanced should still be inherently better than either Standard in several respects.
[/quote]
Thanks Mark,
If I get a request to do a large Ektachrome or Fuji scan project I'll get the Advanced Target.
-
Hello All,
Just read with great interest Mark Segal's recent article on the dark art of scanner profiling.
Actually pretty darn easy and effective. Now profiling a camera, that's a dark art. ;)
An individually measured target is ideal but probably not cost effective for most and I'd prefer a individually measured transparency target than print and that's not easy to do!
-
The reported “0.7” means the average dE(2000) of all the patches in the target has a value of 0.7, which, if correct, is excellent
Curious if you've ever seen a report differ from this because it's exactly what I saw when testing. I have to wonder if it's a 'fixed' value or a colorimetric reality. ::)
-
I raised the same curiosity in the article, because every test I did with all their targets returned the same value.
-
I raised the same curiosity in the article, because every test I did with all their targets returned the same value.
Got an extra target you don't care about and a sharpie pen? ;D
-
That could be fun, but I don't. :)
-
That could be fun, but I don't. :)
Blue painters tape? One or two patches? I'd do it myself but the scanner is back in storage.
-
There are numerous ways to amend a target - but I don't have spare targets and I'm not about to put time into getting them, fudging them and testing; I know what I really wanted to know and share about these processes in other ways.
-
There are numerous ways to amend a target - but I don't have spare targets and I'm not about to put time into getting them, fudging them and testing; I know what I really wanted to know and share about these processes in other ways.
That's fine. But I'm highly suspicious that the report may be bogus. If so, kind of telling for readers of a review, no?
-
Based on the totality of experience I've assembled over the years doing this kind of stuff, I found a series of such identical results to be unusual and I stated my observations in the article. Readers may of course have their own views about the causes, which latter I do not know. Now that you've mentioned it, it is indeed intriguing that both of us obtained the same result using different targets and different scanners at different times between New Mexico and Toronto. Could materials and equipment be that uniform in their behaviour, or is there some other factor at play? LSI's literature on these targets does say that the dE result can be expected to vary generally between 0.8 and 1.8.
-
Based on the totality of experience I've assembled over the years doing this kind of stuff, I found a series of such identical results to be unusual and I stated my observations in the article. Readers may of course have their own views about the causes, which latter I do not know.
IF a target is modified largely with say blue tape and the same dE values are reported, it's a bug. Or they are providing bogus info or both. Simple as that.
I think we're saying to the readers: take the value with a grain of salt. ;)
-
I suspect that if the target were modified as drastically as taping it, the profiling process would breakdown. A thought that crossed my mind when I was looking at all this was to wonder, again speculation, whether under the hood the application is re-iterating the calculations until the result reaches a fixed value - in this case 0.7. I have no idea whether this is likely or even feasible.
-
I suspect that if the target were modified as drastically as taping it, the profiling process would breakdown.
And that behavior alone would be telling. But if it didn't break down or the value did or didn't change, that would be really telling! One patch being covered shouldn't cause such a break down but we'll never know unless I decide to take my scanner out of storage.
-
.............but we'll never know unless I decide to take my scanner out of storage.
If you decide to do so, do let us know how you modified the target and what were the outcomes.
-
Thank you, all, for the informative responses. Though I have a few targets already, I doubt that any individual one was hand measured. I think the batch created target would be good 'nuff for my needs. The difference in price between the batch version and the individually measured version is quite substantial! Maybe they'll put 'em on sale, again... ;)
-
If you decide to do so, do let us know how you modified the target and what were the outcomes.
Here you go. I do see a LOT of tape popping an error (kind of a meaningless error but one none the less), I see still very low dE reports when several patches are covered with blue tape compared to the one report with NO tape also provided below. That makes me think the report isn't very accurate but it's probably an indicator someone did something wrong. Screen captures below show the dE report and what patches were covered so I hope, self explanatory:
-
Wow!
I'm sticking with my printer made patches for profiling my V850. I don't have Silverfast targets but have compared profiles made with the (small) IT8 target using I1Profiler and ones made with a printed target. There is, as expected, little error (ave dE00 .7) when scanning different printed targets but that can be chalked up to using a common spectro for measuring the printed targets together with virtually no metameric error since the scanned target and crosscheck target were both printed with the same printer.
Further, when I scan both a ColorChecker and the larger SG Colorchecker, the average error using the printed target profile is as good or slightly better than what I get with the IT8 but all were in the low 2's. Not sure how much of that is metameric error v difference in instruments since the CC's were measured with an I1Pro 2 while the printed targets were measured with an I1iSis but I expect some metameric error since the CC's are spectrally different.
I might get some Silverfast reflection targets to test. I would think they would be better than the XRite one which was already 9 months old when I got the scanner.
-
Here you go. I do see a LOT of tape popping an error (kind of a meaningless error but one none the less), I see still very low dE reports when several patches are covered with blue tape compared to the one report with NO tape also provided below. That makes me think the report isn't very accurate but it's probably an indicator someone did something wrong. Screen captures below show the dE report and what patches were covered so I hope, self explanatory:
Thanks for doing that Andrew - this is useful. Firstly, it shows that the dE0.7 isn't a fixed outcome, so there is no bug in the software causing it to always report that one value. Secondly, as expected, it shows that when too may patches are taped, the profiling breaks down, which is what I expected. Thirdly, the first two images show that WHERE you place the tape (over the same number of patches) makes a difference to the dE calculation. Not sure why a complete wipe-out of one patch should produce a different error from that of any other. Anyhow, this discussion and your further work on it has produced helpful observations.
-
Thanks for doing that Andrew - this is useful. Firstly, it shows that the dE0.7 isn't a fixed outcome, so there is no bug in the software causing it to always report that one value. Secondly, as expected, it shows that when too may patches are taped, the profiling breaks down, which is what I expected. Thirdly, the first two images show that WHERE you place the tape (over the same number of patches) makes a difference to the dE calculation. Not sure why a complete wipe-out of one patch should produce a different error from that of any other. Anyhow, this discussion and your further work on it has produced helpful observations.
Yes, thanks Andrew. Most interesting.
The change from .7dE to .9dE is about what one would expect if an additional error of, say 100dE from the 4 taped patches was incorporated in the average. It would be interesting to test the profiles made against each other to see if they resulted in large perturbations around those colors. It's quite possible the software discards them as outliers and produces good profiles even while reporting a higher dE. I can easily test the two if Andrew posts them.