Removing Orange Mask from Colour Negative Scans

[peaceman]

Very cool, Carl.
I have basically made the same experiments a while ago... manually in PS, which was quite cumbersome. I realized that I had to find the density vlaues of Y and M new for every film... the results differed big time. Unfortunately I failed in automating it.

In the meantime, I found dedicated software, but haven't really tried it yet... URL for "ColorNeg" is is http://www.colorneg.com/colorneg.html?lang=en .

Video: http://www.colorneg.com/ColorPerfect-vi ... roduction/

Anyway, not intended for MP so far. I recently read that DaVInci Resolve is free now? That should have such capabilities too...

[carllooper]

Thanks Friedemann.

I haven't used DaVinci Resolve but was pleasantly surprised to see on the website for such, that DaVinci Resolve coincidently implements an idea I had just yesterday, where software could automatically recognise a Macbeth Colour Chart in an image stream, and auto-lock onto such to calibrate a transfer.

So there's a version of DaVinci Resolve I found, called DaVinci Resolve Lite, which is free.
Google found me here, under the Australian (au) wing of the website:
http://www.blackmagicdesign.com/au/prod ... olve/color

The download link takes me here to support, where I have to reselect it from a menu
http://www.blackmagicdesign.com/au/support
Select: OS, Product Series = DaVinci Resolve, Product = DaVinci Resolve Lite

Eventually I ended up here where I could download it (after choosing Windows version):
http://www.blackmagicdesign.com/au/supp ... lse&os=win


C

[carllooper]

Ok, so just had a play with DaVinci Resolve.

Not bad. For some painful reason the user interface wants to occupy the entire screen without any apparent way of minimising such? It means to do anything else on the computer requires closing out of Resolve. It's quite painful in that regard. For example, I can't read the tutorial PDF's at the same time as having Resolve open? Perhaps it's something to do with the fullscreen insanity that Windows 8 asks of software: the iMad "touch" look and feel. But even so I couldn't mouse over the top left corner of the screen to switch out of Resolve. Oh well. It was reopening a project quick enough.

In any case the various controls for colour balancing are great. Was able to get a ball park balance for a neg scan in next to no time.

The software requires a CUDA enabled graphics card, which I have, and which eventually worked with Resolve after I updated the drivers on such.

C

[prtprt]

I have also downloaded and installed the free DaVinci Reslove Lite from Blackmagicdesign. A lot to learn how to manage this program...

I also noticed that the program occupies the whole screen. The same in my both computers, one with Win7 and the other with Win8.

I haven't got a solution how to make the Resolve screen smaller, but if you first open the pdf manual, and then open Resolve, you can toggle between the two programs by pressing Alt-tab. It works for me in both Win7 and Win8.

Not the perfect solution, but at least you do not have to close down Resolve to get to the pdf manual.

I do not have a Cuda graphic card, but it works anyway, what I can see now. But I guess it would work better and quicker with Cuda.

/Charlie

[dgalland]

New on this forum I was very interested in this subject. I learned a lot !
I usually invert negatives APS scanned with Nikon Coolscan with a imagemagick script called negfix8
I tried to follow carefully your Photoshop workflow but the result is not satisfactory
Do you have an idea ?
I join (reduced in size) the negative, the negfix8 result (without adjustement) and the result of the workflow.
Thanks for your attention and best regards.
Dominique Galland

[poita]

Ok, so just had a play with DaVinci Resolve.

Not bad. For some painful reason the user interface wants to occupy the entire screen without any apparent way of minimising such? It means to do anything else on the computer requires closing out of Resolve. It's quite painful in that regard. For example, I can't read the tutorial PDF's at the same time as having Resolve open? Perhaps it's something to do with the fullscreen insanity that Windows 8 asks of software: the iMad "touch" look and feel. But even so I couldn't mouse over the top left corner of the screen to switch out of Resolve. Oh well. It was reopening a project quick enough.

In any case the various controls for colour balancing are great. Was able to get a ball park balance for a neg scan in next to no time.

The software requires a CUDA enabled graphics card, which I have, and which eventually worked with Resolve after I updated the drivers on such.

C

Hi Carl, did you come up with a way to remove the Orange Mask using Resolve?

[carllooper]

Hi Carl, did you come up with a way to remove the Orange Mask using Resolve?

No, I didn't pursue any tests with Resolve. But from what I could see Resolve has all the controls you would otherwise need to tweak an optimum inversion. The main reason for exploring orange mask removal was not so much to get good results in the digital domain (although that was one of the reasons) but to prepare the groundwork for how to get equally good results in the analog only domain, ie. in a film to film printing setup.

It was really the computer model I posted that was the aim of the investigation (and the underlying logic of such). The play around in Photoshop, and After Effects, was just a way of testing out the model, before formalising it in code.

If printing from the same stock to the same stock there wouldn't be required any re-balancing because the same stock will automatically cancel out the mask (well it would if one used the same colour temperature light in both cases, ie. during camera exposure and printer exposure). But one doesn't use the same stock when doing a print. Nor the same colour temperature light (in general). And I'm interested in what that means and how one accommodates for that.

In an analog printer the controls one has are just straight RGB level controls, and the use of filters in the printing path, (and whatever colour skews the print stock has). I've yet to pursue this in greater detail but it seems to me, from what I was on about, that for an arbitrary print stock, these controls could be somewhat sub-optimal - that one might need to do some multi-pass printing (with some additional tricks) in order to get an optimum print, and I'm interested in what those might be.

However I assume that actual print stocks (as distinct from some mythical arbitrary one) are created in such a way so that RGB levels and filter controls are all that's required to fine tune a print. Otherwise no print could ever be optimal in a standard setup, and that can't be the case (obviously).

It's more about what to do when working experimentally, with random stocks, and sub-optimal exposures, etc, and how one might "correct" for the eccentricities that each variable introduces. In the first case it's just printing with whatever is available (RGB levels and filters) regardless of any mathematical model, and exploiting the results as one sees fit. Doing bracketed tests and making notes. A mathematical model is just a more refined version of the note taking side of the situation. Identifying patterns in the data and formalising such in some useful equation that might condense the data. The resulting model provides a framework in which one might be able to tweak the various controllable variables further, in some more predictable way, towards either a conventional result, or an exotic one. But getting a good model together can be tough. And it's not in any way necessary. An emperical approach (trial and error) and intuition can give you equally good results, and perhaps even better ones since you are seeing exactly what you get - and if you project such, then regardless of any prior predictability, you are basically saying to an audience: this is not some inferior copy of what I might have otherwise, in my head, wanted you to see - rather it is exactly, grain for grain, what I want you to see. It is by such methods photography itself was born in the first place.

C

[videoguy326]

I've successfully been able to remove the orange mask using Final Cut Pro X. I use the auto color balance filter which removes the orange mask and then use the negative filter to convert the image to a positive.

[carllooper]

I've successfully been able to remove the orange mask using Final Cut Pro X. I use the auto color balance filter which removes the orange mask and then use the negative filter to convert the image to a positive.

Auto colour balance won't remove the orange mask. The results of using auto colour balance might provide you with with a reasonable starting point for obtaining an okay result, but the main discussion in this thread is towards getting a better-than-okay result - towards getting an optimum result - towards getting every last drop of colour out of the original neg as one can.

The orange mask is not a simple orange bias in the image.

If it were such there would be no reason to have it in the neg in the first place, and it would be very simple to remove, and auto colour balance (being a very simple algorithm) would provide a result very close to an optimum end result.

But the mask is actually an image, indeed it is made of two images, one yellow and the other magenta. To "remove the orange mask" is to extract these images and use them to rearrange the data so that one has the otherwise encoded RGB colour signals back in the correct channels. After doing that one is in a much better position to get the best possible colour out of the image.

The reason for the orange mask is that the dyes in film are not optimal dyes. The cyan and magenta dyes (so called) are not ideal. The cyan dye is bluer than it should be, and the magenta dye is redder than it should be. Yellow dye is the least offensive. This bias in the dyes means they would be controlling more than the channel of light they should be, leading to a pollution of the original colours (and corresponding target colours) that it otherwise encodes. The result is a less than ideal colour.

In reversal stock, for example, one might note that the colours in such are never quite as vibrant and diverse as one would like. Reversal stock has to put up with the pollution. It does the best it can, but it can't do any better.

A bit like using auto colour balance. While the result is okay it's not as good as it could be.

The masks in neg film are designed to stop the cross pollution of colour information that less than ideal dyes would otherwise produce. The original colour is effectively encrypted in the film, in a way that compensates for the dye inadequacies. Decoding the data becomes a little bit more complicated, but I think the results of the extra effort in decoding it do make it worthwhile. But yes, it's a bit complicated.

For a good explanation of the orange mask I fully recommend this one:

http://www.brianpritchard.com/why_colou ... orange.htm


C

[djarwood]

Hi all,

I did a plugin to do the filter removing job, the source code is here:
https://github.com/edubois/kaliscope

Kaliscope is a software I am making for telecinema devices. It is based on OFX plugins, and it includes a plugin called filterCleaner that is used to remove orange mask. You only need to know (or guess) the color of the orange filter.

If you are interested, I will post link to a mac os compatible plugin very soon.

[djarwood]

Hi all,

I did a plugin to remove orange filter (or other kind of color), you can download it here:

http://crearty.fr/djarwood/ofxPlugins/d ... dle.tar.gz

Please tell me if you have a problem using it.

[Dennis_xuejing]

Hi Carl,
It is wonderful process.

Actually removal of orange mask is really painful to me and I am really struggling how to properly get rid of the mask and get the right color.

Before I came here I found another process which was come up with 15 years ago by Burton in the Adobe forum, you can check it by clicking the link as follow: http://www.dpreview.com/forums/post/499854

To me looks both process have same theory basis and similar logic, but the outcome is the different. Would you help to check and let us know if there is any difference between two processes?

Another question is that different film has different orange mask, is it possible that we can measure and quantify the mask and then remove the specific yellow and Magenta image from yellow and magenta channel (in the Burton’s process he said he would develop a updated process with measure of the mask, but unfortunately I cannot find the update) . Right now in your process to magenta mask and yellow mask you will adjust the output level to 51(20% of 255). To the different orange mask the number shall be different.

Your reply and explanation will be highly appreciated.

Dennis

[djarwood]

It's looking like people doesn't care, but I'm trying again: I did an OFX plugin compatible with many hosts (DaVinci resolve, Nuke, Sony Vegas, ...) that solves this annoying problem.

If you want to try it with my homemade host, here is a link on the sourceforge project:
https://sourceforge.net/projects/kaliscope/

[carllooper]

Hi Carl,
It is wonderful process.

Actually removal of orange mask is really painful to me and I am really struggling how to properly get rid of the mask and get the right color.

Before I came here I found another process which was come up with 15 years ago by Burton in the Adobe forum, you can check it by clicking the link as follow: http://www.dpreview.com/forums/post/499854

To me looks both process have same theory basis and similar logic, but the outcome is the different. Would you help to check and let us know if there is any difference between two processes?

Another question is that different film has different orange mask, is it possible that we can measure and quantify the mask and then remove the specific yellow and Magenta image from yellow and magenta channel (in the Burton’s process he said he would develop a updated process with measure of the mask, but unfortunately I cannot find the update) . Right now in your process to magenta mask and yellow mask you will adjust the output level to 51(20% of 255). To the different orange mask the number shall be different.

Your reply and explanation will be highly appreciated.

Dennis

Yes, that seems to be more or less the exact same procedure I was proposing. Different stocks will use slightly different dyes so the various percentages involved will differ. A guesstimate on values will often be quite fine. One still has to do a traditional colour balance afterwards anyway as there will be biases in the light used during capture, biases in the camera used to do the capture, biases in the file formats used (different log mappings) and so on. Properly speaking one should calibrate the entire workflow, eg using colour charts, exposed on film, and taken through the digitisation process and compared to the original chart. But just a rough demasking and a good eye during subsequent manual colour balancing will get better results than not doing so.

If one just uses a flat subtraction of some orange value, (as many do) all one ends up with is something the same as having shot reversal - which in many situations would be just fine - a lot of people - including myself - have shot reversal and found it quite satisfying. But negative provides some extra colour that reversal can't. So properly demasking the negative gives us that extra colour that reversal can't give us.

C

[Hansha]

One of the differences between Burtons and this procedure is that he assigns 400% to total ink limit, compensation for the Black 100% that are gone. Also, he makes the standard dot gain -10.

Of course, he also works with opacity, not levels. I have far too little knowledge to judge what works better but sure would like to know.

The other question that bugs me are the differences in the masks and whether the data for correct subtraction could be derived from a simple analysis of the non-exposed part of the film? We are now dealing only with pure CMY removal. Is the orange cast really made up just from this two inks or does the layer include some of its own chemical hue?