Hello, I tried you plugin with vuescan and opticfilm120.
Good,
I have silverfast 35mm IT8 target.
Oh my.... The profiling question. First off, in the past 90% of the people asking this never write again after my response. They either don't understand my answer or don't like it. So the short answer is what is referred to as "profiling" was nonsense all along. It cannot work. It defies physics and is only there because engineers didn't know what else to do. For starters I'll forward a message below I wrote to someone a while ago.
I followed the linear scan tutorial here:
http://www.colorneg.com/scanning-slides-and-negatives/scans/Hamrick-Software/VueScan/Good, there is no IT8 nonsense in that workflow.
However how do I create the ICC profile from the image scanned this way???
You don't.
But the problem is that resultant profile distorts gamma, it tires to lighten things up as there is no way to specify linear gamma. The scans obviously are very dark at linear gamma.
The "profiles" created that way will always distort various things in uncontrollable ways. Let me pass on a message I wrote a while back...
This is the exerpt from the message I intended to forward. It deals with the notion of "assigning a scanner profile" to a negative.
The person I wrote it for decided to ignore it and to advocate his troubled workflow anyway.
-----------------------------------------------------------------------
In summary, what you recommend doing is more often than not diametrically opposed to what we recommend on our web pages and the help system. Both the idea of being able to eliminate the scanner as a factor from your imaging results and your implementation of that using scanner profiles are seriously flawed.
Color accuracy is mathematically limited in any three-primary imaging system but beyond that limit what you need is not so much color accuracy as consistency of the color flow in an image with the laws of physics that the eye has been developed by nature to expect. When treated correctly, color film images in the analog world naturally follow these physical laws and so produce pleasing results when printed right. The current treatment of digital images departs from these physical laws and so produces images that don't always look bad, exactly, they just don't look quite right - sometimes by a long shot.
[...]
So for now let me just give you one example of something that actually happens in what you do. You create a scanner profile using an IT8-Target (a core item of the physically nonsensical technologies we are trying to overcome). Key component in scanner profiles are color lookup tables. They’re typically 32x32x32 values in size. That means that 32.768 individual RGB color combinations can technically be assigned from any input to any output. Completely ignoring all physical relations between such values. IT8-derived profiles just warp colors around more or less intensely and in a way light never could.
What a scanner does is capture three individual intensity signals. Each represents a weighted average from the spectrum of visible light. A collection of red, yellow, orange and rose colors might have an identical red component captured by the scanner. The same might apply for a collection of blue, green, brown and turquoise colors. I quickly illustrated that for you by simulating two such groups of colors - they really could exist this way:
[]
Looking at the red channel note that there are only two different levels of red intensity:
[]
What an IT-8 derived profile might do nonetheless is things like making the yellow brighter which would require its red component to be increased and red darker which would require the opposite of the same red intensity reading. Rose, and orange would be adjusted differently again.
Can you see that such is physically impossible in terms of light? The only thing you can do with CC filters on an enlarger for example are linear transforms. That means all pixels in the entire image must change equally or not at all. Filter dialed in or out. The kind of force matching technology used in digital "profiling" will make a given target come out precisely as the intended target alright because all target values are known but for any other picture it’s pretty much moot. But it is neat that the succes of such systems is measured based on the ability to render defined targets...
But wait you’ll think: This is crazy, my workflow does work… Please consider this example carefully:
For negatives you create a linear scan. That scan is basically RAW. You assign a scanner profile to that scan and it does what I outlined above (warp intensities that should stay identical physically in opposite directions etc.). That happens for the yellow, orange, red, rose group of colors I envisioned and the same for many other such groups.
Now the essential bit of information here is WHEN does that transform get applied to the data and HOW. It gets applied by Photoshop before the data is displayed. ColorPerfect is completely ignorant of that. It uses the linear RAW scan and does its work. Photoshop will display the preview after each step and warp it according to the original scanner profile. The image changes, the profile stays the same. Funnily in the end what was orange in the negative now no longer is because the orange mask got canceled out. The image was inverted so what was green is now magenta and vice versa what was red is now cyan and vice versa what was blue is yellow and vice versa, brightnesses have changed all over the image and that in the most severe way imaginable and yet after all of it Photoshop still applies the same IT8-Based scanner profile which says red should warp this way for this scanner, yellow should warp that way etc. There is NO correlation to the scanner in that transform whatsoever at this point. It’s a completely arbitrary color shift.
ColorPerfect works with the assigned scanner profile because it gets the unaffected linear data. If you were to “correct” the scan for any effects of the scanner you’d have to convert to your working profile before applying ColorPerfect – that will wreck everything though because it destroys the physical relations ColorPerfect is built on and so ColorPerfect will fail after attempting such.
Lookup tables are completely unsuitable for input devices. Once color integrity has been produced and the final image looks the way we want it to look profiling can be quite valuable. That is for output devices such as screens, printers etc. It no longer is as important to adhere to the physics of light in that phase. In it we just want to keep our result to look as close to what we created as possible. Any means that do work in that are fine. The changes involved will be subtle anyway - while the changes during capture and processing can be tremendous.
This is really just the beginning, you seem to use none of the tools we offer that are superior over even such seemingly simple things as the shadow and middle gray sliders in Photoshop but I’ll leave that for later. For us to continue on this road you must be prepared to throw over most of your workflow. If you are I can assure you that over time I can probably give you what you need to truly understand your imaging process. If you aren’t this is all tilting at windmills.
Best regards
Christoph Oldendorf
