Removing Orange Mask from Colour Negative Scans

[carllooper]

I had this in an unrelated thread, so thought I'd put it here in it's own new thread.

It's for anyone who has tried scanning colour negative themselves on their own DIY scanner, only to find it seemingly impossible to get a good colour result out of the scan. The reason it's so difficult is that the orange mask in the negative is not a flat signal. It actually contains an image (in fact two images) due to the way masks and dyes are coupled. These images need to be removed. So it requires a little more than just patiently moving colour balance sliders around all day. However it's relatively simple to implement in any NLE, once you get your head around it.

THEORY
A colour negative consist of 5 layers:

1. Blue Sensitive Layer (becomes Yellow Dye)
2. Yellow Mask (stops blue light diving deeper into the emulsion, but lets red and green light pass deeper)
3. Green Sensitive Layer (becomes Magenta Dye)
4. Magenta Mask (stops green light descending deeper into the emulsion, and lets red light pass deeper (and would also let blue light pass but there is no more blue light at this stage)
5. Red Sensitive Layer (becomes Cyan Dye)

In addition to this, each masking layer (Yellow Mask and Magenta Mask) is coupled to the layer directly below it: so the Yellow Mask is coupled to the Magenta Dye layer below it, and the Magenta Mask is coupled to the Cyan Dye layer below it. The result (during film development) is that the Yellow Mask acquires an inverse image of the Magenta Dye layer, and the Magenta Mask acquires an inverse image of the Cyan Dye layer. Both of these images need to be removed from the digital scan. But how? Well there is no Cyan Mask in the film which means the Cyan channel of the film is unpolluted. Since the Magenta Mask in the film is an inverted image of the Cyan layer, one can recreate the Magenta Mask from a copy of the Cyan layer, and then subtract that from the Magenta channel! The result being an unpolluted Magenta channel, ie. without the Magenta Mask. Repeat the same logic for the Yellow Mask and one has then removed the Yellow Mask, ending up with an unpolluted image. The rest is just fine tuning the image in the way you would otherwise fine tune an ordinary colour image.

PRACTICE

1. Digitise the negative with a blue filter such as an 80A.
The purpose of this is not colour correction. It's just to ensure the peaks in your RGB signal are closer together than they otherwise might be. It's for optimising the limited bandwidth of your capture device. A downside is that it will reduce your light source by about 2 stops, so you might want to skip this step.

2. Convert RGB capture to CMY (no K)
It's possible to work entirely in RGB space but it makes more sense (from an understanding point of view) to work in CMY space (the same space in which film works). You want to make three separate channels, each of which, on their own, are black and white. For example, in Adobe After Effects, to create a Cyan channel, you could use the Channel>Shift Channels effect on your source, setting: Take Red From: Red, Take Green From Red, Take Blue From Red. But why from Red? Well Cyan is the inverse of Red (Cyan = 1 - Red) so we start with red, however we won't do the invert here. We'll do the inversion later. In any case use the same logic to obtain a Magenta (from Green) and a Yellow (from Blue).

3. Copy Cyan channel into a new channel called Magenta Mask.
The Cyan channel is the only channel that contains a pure signal (uncontaminated by any mask) so we don't want to change this channel in any way. We want to leave it as is. What we are doing here is creating an image of what the Magenta Mask, in the film, would look like. We're effectively mimicking the way in which the Magenta Mask, in the film itself, is created.

4. Reduce levels on Magenta Mask to about 20%
This is to emulate the relative strength of the Magenta Mask with respect to the Magenta Dye layer, ie. it's about 20%.

5. Subtract Magenta Mask from the inverse of the original Magenta channel. Invert the result and call it Magenta Repaired.
Here is where we are correcting the Magenta channel. Without an image of the Magenta Mask, we wouldn't be able to correct the Magenta channel. We'd otherwise be fiddling with CC sliders all day long without getting any closer to a result. But here we have the magenta channel repaired. Magic.

6. Copy Magenta Repaired to a new channel called Yellow Mask. Reduce levels to about 20%
This is the same logic as Step 5 but instead of using the original Magenta channel to create an image of the Yellow Mask, we're using the repaired Magenta channel (this is in keeping with how the film develops. The Yellow Mask is coupled to the Magenta Dye image but not to the Magenta Mask)

7. Subtract Yellow Mask from the inverse of the Yellow Channel. Invert the result and call it Yellow Repaired.
We now have a repaired Yellow.

The following just recomposes our repaired CMY images back into a single RGB image.

8. Remove Green and Blue from the Cyan Channel, to make a new result called Red
9. Remove Red and Blue from the Magenta Repaired to make a new result called Green
10. Remove Red and Green from the Yellow Repaired to make a new result called Blue.
11. Combine Red Green and Blue with blend mode = Screen.

13. Invert the result of the above to obtain a positive image.

You now have an image in which the orange mask (yellow and magenta masks) have been removed. Feel free to dance around the room. From here on in it's just ordinary colour correction using your eyeballs and creativity.

If anyone needs any clarification, please ask.

Carl

ps. for the reverse process: digital out to print stock, the reverse process would be required - ie. ensuring your digital image has the orange mask put back in (plus some extra tweaking of the RGB channels to accommodate bias of print stocks).

[sportique]

for me the trick seems to be to set the white balance on the scanner to neutralize the orange cast on the negative and then invert the colors and adjust the green and blue curves in After Effects.
I did some tests with re-scanning some neg I had previously had transferred on a FlashScan and was able to get the exact same result

[carllooper]

You'll find your scans will be a thousand times more colourful if you follow the procedure outlined in the post.

Simply adjusting the white balance and/or colour channel levels won't cancel out the pollution from the orange mask. This is because the orange mask is not just a simple colour bias in any of the colour channels. The orange mask is actually composed of two images (inversely proportional to the cyan and magenta records), and these two images can not be cancelled out by just adjusting colour balance (or curves, or gamma etc). The only way of properly cancelling out the mask is by reconstructing the images they contain, and then subtracting out those images from the magenta/yellow record. The cyan record, insofar as it is unpolluted by any mask, provides the key to the recreation of such a counter mask.

Following this procedure you will be pleasantly surprised by how rich and vivid the colours in your negative actually are.

C

EDIT

Here I've reverse engineered one of the frames, by inverting it back into a negative, and then taking it through the prescribed orange mask removal process, before inverting it back into a positive again, followed by an otherwise ordinary colour balance and gamma correction.

[beaunizo]

Any chance of a plug-un or dedicated program which does work using this Looper method on a set of scanned imaged? Still or ciné

[supa8]

Does this mean that when scanning with HD Moviestuff units, Sniper Pro HD for instance, you would be better off scanning the film as is, without adjusting the white balance to have something bluer? then adjust it in After FX or other software?
How about Colorista II from the RedGiant Bullet Suite?
Would this be suitable to do the procedure you mention above? It might even be more efficient? (Colorista II takes advantage of cuda/GPU acceleration features + no need to open in another program as the plugin can integrated within Premiere, FCP etc).
Then, for footage that has been scanned previously with the white balance reset to tame the orange cast, then inverted via Moviestuff's program, how can one best grade the material to obtain better quality results?

[carllooper]

Yes, it's better just to get as raw a scan as possible.

It's not completely necessary to do any light source filtering during the scan. The only reason you might do so is to nudge each of the colour signals closer together, so that they all fit within the bandwidth of the sensor being used. Without such a filter the toe or shoulders of one or more colour components might clip (because they are spread out too far from each other), but otherwise the bulk of the signal will probably be fine. Either way, doing a white balance is a very good idea. It will keep each of the colours at least within the best possible proximity of each other.

Once the scan is done, if the technique used to tame an orange mask was just an ordinary colour balance (without too much lossy fiddling around such as crushing blacks etc), most of the original information will still be in the result. So one can just invert it, ie. back into something looking like the original negative (and just ignore the colour balance there as one will correct that later), and then perform the prescribed orange mask removal process, re-invert the result back to positive, and then do ordinary colour correction / gamma correction to correct for any ordinary bias in the outcome.

I'll have a think about writing a plugin (always depends on time). But you can manually do it using standard effects available in NLEs (After Effects, Premiere, etc) or Photoshop for that matter. I'll see if I can write up a tutorial for how you would do it in Photoshop, and then you could transpose that into After Effects or Premiere or any other NLE.

C

[supa8]

Great
Thanks a lot
I definitely would be happy to throw a few bobs towards some assistance with this process.
If I had some presets in Premiere Pro CS5.5, that I could save an easily reuse and tweak, it would save me a lot of time and hassle, on top of getting better results for neg film.

[Nicholas Kovats]

Great analysis and practical methodology once again, Carl!

[carllooper]

So here's an example of doing the orange mask removal in Photoshop.

For this to work you need to convert your image into a CMYK image without the K. You can do this by creating a custom CMYK profile in which the Black Ink limit is set to 0%. This will interpret your digital image in the same way that colour negative is created, ie. without a black channel.

So after loading a test image, select:

Edit > Convert To Profile ...
Profile: Custom CMYK

And then enter the following:

Ink Colors: SWOP (Coated)
Dot Gain: 0%
Separation Type: UCR
Black Ink Limit: 0%
Total Ink Limit: 300%




And then follow this:




Something similar can be done in After Effects, but (as far as I can tell) After Effects doesn't provide CMYK space in which to work, but it's easy enough to prepare comps that can do it. I've indicated one such approach in the first post, using Effect>Channel>Shift Channels.

If I get time I'll post an After Effects tutorial.

[milesandjules]

Interesting process Carl…..you certainly know your colour stuff.

I wonder how does fancy digital telecine suites deal with the orange mask in realtime?

I want to try this technique with neg footage from the retro 8.

Carl...Any chance of you showing how to do it in after effects?…I'm hopeless at trying to find cmyk colour settings in it.

[Tscan]

Interesting process Carl…..you certainly know your colour stuff.

I wonder how does fancy digital telecine suites deal with the orange mask in realtime?

I want to try this technique with neg footage from the retro 8.

Carl...Any chance of you showing how to do it in after effects?…I'm hopeless at trying to find cmyk colour settings in it.

The Retro8 can do an invert during capture. The results are pretty good but you still have to do a few adjustments. I'm curious if Carl's method would produce better results easier? At least if I could understand them, because his stuff is often way over my head.

[milesandjules]

yeah same here :oops: Be interesting to see the difference between retro 8 neg conversion and just scanning it as a negative than using Carls technique.

[carllooper]

Here's an example of the difference the orange mask removal provides.

The top left image is a scan of the original negative as found here on a photography forum looking at the same problem: http://www.lf-photo.org.uk/forum/viewtopic.php?t=2120
The bottom left image is the best that the photographer was able to achieve using only colour balancing methods (as elaborated in that forum).

The top right image is the result of putting the original negative (top left image) through the orange mask removal process.
And the bottom right image is then just a colour balance and gamma adjustment on the top right image. I've opted for a darker look.

[carllooper]

The theory section in my original post is somewhat misleading if not just plain incorrect. I'm still learning this stuff myself.

However the practice side of the post remains the same (otherwise the images I've posted wouldn't look so good).

THEORY TAKE 2

So the components of the mask are not in layers separate from the dye forming layers. Nor is it the principle purpose of the mask to stop certain colours of light from diving deeper into the emulsion. The mask is unused silver-dye couplers (which otherwise participate in the creation of a dye image from the silver image) that are visible in the completed film image. This is intentional. Couplers in film can be either colourless (as they are in a "maskless" film, such as reversal) or have a colour in themselves (as in colour negative film). Colour negative film is designed for printing rather than observation in itself, and therefore it doesn't matter what the negative itself looks like, but how it would best work when printing to print stock. The idea of giving couplers a colour is to compensate for impurities in dyes. Print stock will have the same issue with dye impurities as camera stock, but in relation to the light signal being mediated, it will be different dyes that will have this problem (because the print will be coding an inverted colour image of the negative). What this means is that a film manufacturer can compensate for dye impurities by taking advantage of the printing process: from negative to print. During this process the problem of dye impurities can be cancelled out. It so happens that the best way to solve the problem of dye impurities is to give the couplers a certain colour or what we otherwise call the "orange mask".

The result of this strategy (which is towards the design of a perfect negative film) is that the couplers for the magenta layer won't be made clear, but will be made a little yellow, and the couplers for the cyan layer will be made a little magenta. The coupler for the yellow layer is made clear. During development couplers end up with a value that is in inverse proportion to the dyes they otherwise help to facilitate. Where there is more dye colour in a particular layer there will be less of the coupler colour and vice versa.

From this it should be clear why a maskless negative film would not be a good idea. The entire purpose of the mask is to facilitate better colours in the end result (in the print). It should also be clear why reversal film can't have a mask, because there is no opportunity during the reversal process to perform any cancellation of any dye impurity compensation. It also explains why the colours in reversal are not as pure as the colours obtainable from a negative (ignoring the voodoo that was Kodachrome 40).

But while this all works just perfectly for film to film printing, it's awkward for film to digital printing. However the mask still facilitates it's purpose in this context. All it means is that the digital print needs to emulate the way print film behaves. And that is what the "orange mask removal" process is effectively doing. It is digitally emulating the way a film print otherwise resolves the same outcome.


Anyway, have a look at the last diagram, on the last page, in this publication, for a very good idea of the final component images in a colour negative. Its really only this one needs to work out (from scratch) what is to be removed in an orange mask removal process. What is to be removed is the yellow coupler image in the middle magenta dye layer, and the magenta coupler image in the bottom cyan dye layer. Insofar as there is no cyan coloured coupler in the film, it means the cyan signal in the film is pure (ie. not a mixture of a dye image and a coupler image), and it becomes the basis for a reconstruction of the coupler images (ie. the orange mask) otherwise unseparateable from the yellow and magenta signals.

http://motion.kodak.com/motion/uploaded ... _films.pdf

C

[S8 Booster]

this method is simple to use and the results are great:

http://photosolve.com/main/resources/pw ... index.html Method B

works well for me in stills.