Flextight Colour Neg Scanning

[Ted Baker]

The books you point are really interesting, but those explains (amazing) how color was managed by analog means.

Giorgianni was the Designer of Kodak's PhotoCD so no its not just about analogue, I think your missing a few parts to the puzzle and you would enjoy that book and it would assist you greatly.

Let me point a case, Star Wars 7 and 8 was shot in film, and the first thing they did was scanning the film, they made the digital process, and finally they also printed the film for the IMAX release. All the digital process was made in a RGB space, ending in a printed film that had an spectral complexity like the negative had, but nothing was lost when the spectral information of the Vision 3 negative was reduced to 3 numbers. Any lost information was lost at the exposure time. They used V3 because it was taking the information they liked.

Star wars like most films shot of film in the last twenty years was scanned using Kodak Cineon or DPX system. This is something I understand well, as it is one of the few processes that is very well documented. A cineon scanner records what is referred to as print density, it uses 10bit per channel to record the density, the density measurements are calibrated as if the negative was printed on Vision 2 print stock (the stock that existed at the time) resulting in neutral grey on the print stock which can be measure using status-a measurements. These density measurements are not the same as status-m or typical measurements of the negative, but what the resultant print density would be. It was designed originally as an analogue to analogue system, with a digital intermediate.

It was from these 10bit density measurements that a CLUT can used to convert to digital projection, or digital effects can be mixed in and the whole thing output back to 10bit density measurements and printed on print stock.

I understand the concept of mapping but also what I what I was trying to point out is that you need more than an IT8 target (which is not designed for this purpose) to do it properly.

My own solution is based on the cineon documentation, and I am currently attempting to use a macbeth color checker photographed and printed on RA-4 paper and measured with a colour meter as well comparisons with fuji frontier, and noritsu scans to calibrate my own output.

[Pere Casals]

Well, an IT8 target calibrates the scanner job, what you do is clearly beyond it, I understand that. You are replicating the color analog processing for the hybrid, challenging task.

What I'm pointing is that this task should be possible with a 3D lut for each film/paper combo.

[Pere Casals]

...so I'd use 3D LUT Creator to map negative colors (ITU calibrated) directly to the corresponding print colors(also IT8) . This should work, but a particular 3D LUT should be used for each negative/paper types...

The generated 3D LUT would contain what the IT8 calibration cannot do...

[Ted Baker]

...so I'd use 3D LUT Creator to map negative colors (ITU calibrated) directly to the corresponding print colors(also IT8) . This should work, but a particular 3D LUT should be used for each negative/paper types...

The generated 3D LUT would contain what the IT8 calibration cannot do...

Putting aside that it8 targets made on negative stock don't exist. (You could make your own with a colormeter). How does your process account for the print exposure?

In reality It's a multistage process, what your starting to describe is now more involved than you first suggested ;-)

But your on the right track for sure.

[Pere Casals]

Putting aside that it8 targets made on negative stock don't exist.

IMHO this is irrelevant, common scanners work nearly the same, so we have a consistent starting point with the negative scanned image and a consistent end point with the print scan.

Process can be calibrated from reference prints that come from a range of practical situations.

Of course we can also make reference negatives/prints by using color checkers, with different exposures.

How does your process account for the print exposure?

IMHO the conversion should be assisted, as the analog printing process is. The calibration of the conversion can be done with a negative exposed in standard conditions and printed optimally, but operative conversions can be assisted with user adjusting exposure and CMY color balance as it is done in the darkroom. Those adjustments would modify the negative image before being the 3D LUT applied.

In reality It's a multistage process, what your starting to describe is now more involved than you first suggested ;-)

I reiterate what said in the first suggestion, each stage can be mapped with a single 3D LUT, but a number of sequential 3D LUTs can be combined in a single 3D LUT, so my suggestion is considering the full process is a black box and then we only need to find a 3D LUT working as the transfer funtion of the back box, this is generating the LUT that takes the transmission sprectrum of the reference negative and delivers the scanned reference print.

This is straight, today we have tools that generates a 3D conversion LUT if we have the input image and the output result we want.

[interneg]

My own solution is based on the cineon documentation, and I am currently attempting to use a macbeth color checker photographed and printed on RA-4 paper and measured with a colour meter as well comparisons with fuji frontier, and noritsu scans to calibrate my own output.

You've also got to remember that cinema print stock is colour timed for a xenon arc lamp, not daylight...

More importantly, are you making those RA4's optically or via scanning? Both of those minilab scanners (from my own experience matching Flextight scans to them and to optically printed darkroom RA4's) significantly cripple certain colours, especially at the saturated ends of the blue/ yellow continuum ('b' in LAB) - probably for perfectly reasonable production related reasons which may have more to do with sociological definitions of 'what a colour negative print should look like' and minimising chances of significant user error by a semi-skilled operator - after all there is a big price difference between a master darkroom print & a print off a Frontier. That said, these machines can do a pretty excellent job of getting rid of the mask correctly, though the subsequent 'looks' they bake-in may have more to do with what 'looks like it could be correct' rather than what the film actually has to offer. The density & colour is thence internally altered to minimise the chances of mismatch with the output media, no matter what the output media's possible gamut is - indeed, I recall that there's there quite a lot of research going all the way back to the early days of mass market colour printing which effectively states that people will often accept quite a range of colour, as long as the density is correct.

I get the sense that you are making this exercise significantly more complex than it needs to be - possibly because you are struggling with compromised scans. Fundamentally, it's relatively simple (within the context, making the film & paper emulsions work correctly together was drastically more difficult - hence the need for the mask in the first place!): the software has to remove the mask in much the same way as chromogenic paper would 'see' the negative, and from then onwards it's up to someone skilled at colour correction (which is not a problem in cinema or higher end photography). And it's that problem of skill & the cost thereof that leads to tightly calibrated compromises in systems aimed at reducing the chances of gross operator error.

As to Pere's comments about the 'looks' from Noritsu/ Frontiers, even with a limited range of choices (say 5-7) for perhaps 5-7 controls, you very rapidly end up with a couple of thousand possible combinations, very few of which are desperately difficult to reproduce. Indeed, they often seem to have more in common with the tendency to colour grade cinema to intentional colour casts, such that when a 'straight' print/ inversion of the neg is made (respecting the colour balance of the film) it actually looks odd because we've been so conditioned to expect 'off' colours.

[calebarchie]

Interneg please, I am enjoying this pissing match in every scanning related thread

[Ted Baker]

IMHO this is irrelevant, common scanners work nearly the same, so we have a consistent starting point with the negative scanned image and a consistent end point with the print scan.

I am sure you can use them, just like you can use 400iso film for other than "sports and action". Maybe an Ektachrome target has the same dyes or close enough to Kodak negative stock. But perhaps others are interested in how they are supposed to work, I certainly am.

I reiterate what said in the first suggestion, each stage can be mapped with a single 3D LUT, but a number of sequential 3D LUTs can be combined in a single 3D LUT, so my suggestion is considering the full process is a black box and then we only need to find a 3D LUT working as the transfer funtion of the back box, this is generating the LUT that takes the transmission sprectrum of the reference negative and delivers the scanned reference print.

I agree, and it is a very useful technique to compute a combined LUT for multiple stages for each value in your input range i.e. make 65536 calculations instead of 50 million, for 50mexapixel scan. If I misinterpreted your first post then sorry, it seemed a little simpler than the task at hand. It is also however useful to bring the image into a linear colorspace, if you intend to use a modern editor even if the output is not expected to be thus.

This is straight, today we have tools that generates a 3D conversion LUT if we have the input image and the output result we want.

If you mean matching the specific output from scanner/software combination if you have access to the same negative. i.e. Someone gives you the finished file from an "acmetight" scanner processed with "acmeshop", and they give you the negative that is straightforward. But that is just fancy copying.

You've also got to remember that cinema print stock is colour timed for a xenon arc lamp, not daylight...

And they even turn the lights off when you view it... Yes the print stock (and camera) is definitely engineered differently to make the most of different viewing environment, but a lot of the technology is the same, plus it is well documented and there is more than one vendor that makes the equipment.

More importantly, are you making those RA4's optically or via scanning?

No printed myself, borrowed a meter, will buy one later. The noritsu and frontier scans you have everything in the file. At some point I would like to try the flexcolor software to see how it works. I was disappointed to hear remarks on this thread that the flexcolor software wasn't all that special with regard to negative processing.

I get the sense that you are making this exercise significantly more complex than it needs to be - possibly because you are struggling with compromised scans.

Perhaps but like many carpenters, machinists etc who are build there own view cameras, while there are perfectly serviceable cameras available, they have a reason for doing so. I am currently interested in image processing.

[Pere Casals]

You've also got to remember that cinema print stock is colour timed for a xenon arc lamp, not daylight...

Interneg, this is true, but that problem it's very easy to overcome. As the xenon arc lamps have a well continuous 6200K (around) color temperature spectrum... the problem can be addressed with a plain color temperature correction after the calibration stage that is not to degradate the information quality at all.