Removing a Magenta Cast from a 35mm Print

Mazin Power Sep 12, 2013

  1. I'm working on a project where we're restoring a 35mm print that's gone totally magenta.

    Our process is to scan on an Imagica to 4k DPX and then grade in DaVinci Resolve.

    Images of the film and scopes can be found here:

    I've managed to do a fix by separating the RGB channels and adjusting the red channel, then doing a basic balance. The results are OK, but I feel I can do better. Just wondering if anybody has suggestions for an approach they would take? I'm not entirely happy with the one I'm currently on.

    For the curious, here is a DaVinci Powergrade file:

    And here is one of the DPX frames (40mb):
  2. Curve tools, down with the red and up with the green and blue shape the curves to make it look good.

  3. 1.jpg 2.jpg
    Another CC for you
  4. Amazing what can be done with a RGB data scan and these tools.

  5. Some very humble tools here, nothing near your Resolve 9 Spaceship like panels.

    A second floor public library, a 2009 iMac, a keyboard, a mouse and Xcode for my on-image adjustment tool to set the black, grey and white points.

  6. You don't need a panel for everything, certainly not basic restoration like this, did you color correct it clip to clip? Usually with a timed print I find I can fix the basic fading and more or less find the color timing from the print, home movies are another thing....

  7. Hello Mazin

    Here is a cc with SCRATCH.
    I'm just using Gain, lift, shadows, gamma and highlight...
    And a HSL qualification in the skin tone to change Hue.

    This is the fist pass. We can arrange the sky and the tree shadows for instance...

    Construct 2_Construct 2_0002.jpg

    Attached Files:

  8. Thanks Rob, Paul and Burt... all very helpful!
  9. Hi Mazon,

    Interesting post.
    Welcome to my world, color restoration.
    I'm Senior Colorist for a South Florida company that is restoring 35mm classic films for broadcast and blu-ray.
    To date, I have completed color restoring nearly 200 films, color and black/white.

    Although a custom curve is a tool you should know and use, it is only one way to adjust the image.
    I might suggest, instead, that you pull a luma key of the sky, and while you're at it you may want to adjust the hue/sat of the foliage.
    Certainly you can re-image the color of the grass, to bring it more in line with our muscle memory of what grass looks like.

    Most of the light in the picture is in the mids and high end.
    It's up to you, but I would leave the custom curve until the end.
    And when you do apply it, I would do a gentle roll off on the curve and push the lift just a bit.
    I might suggest using your soft curves to see what, if any, effect they will have.

    Just sayin'.

    Best, J
    Robert Houllahan likes this.
  10. Took a shot at it while rendering... Im not a restoration expert, but I like challenges.


    Attached Files:

  11. Jake are you grading from Data? What I have been doing is a bit based on having a 4K data scanner and being the person who sets up the scan. The curves workflow is just what I have come up with as I have gone along, I unlink the curves and balance R,G,B and L with them in the first node and then add additional nodes for hue/sat etc. Would love to hear more thoughts on better ways to do this.

  12. I would love to see James' version, a real professional versionl. Thanks

    2 copy.jpg
  13. This was a quick shot at it (my GF wants to go out ans is staring) pretty happy with the sky and trees the grass needs some work...

    Attached Files:

  14. James, I'm curious - why do you suggest keys and secondaries over primary adjustments? My thinking on this is that since theres a global problem with the print, it should be a primary adjustment / colour channel adjustment that brings it back in line.

    These are all really interesting approaches. I'm gonna have to check all these out when I get back to the office on Monday. It's pretty clear that there's a number of ways to go.

    I'm in inclined to judge the green based on the colour of the trees and obviously blue based on the sky... I think the grass is meant to be ruddy brown... must have been sun-burned.

    By the way I love the new feature in Resolve 10 that lets you selectively grade RGB and YUV colour channels in a node!
  15. Hi,
    I used also lumakey for sky and grass, do not use chromakey in resolve at all, scratch is much better in that, but for restoration I wouldnt use chromakey. And used hue shift for grass at the end. RGB offsets done most of work. As James sad a custom curve is a tool you should know and use I dont have a glue how to handle it right way yet.

    Attached Files:

  16. or more neutral and less contrast..

    Attached Files:

  17. I really like that one Gabor, can you share some of your technique?
    James, if you could take a crack at it and walk us through a little bit as to how you achieved the result, there is little out there as references on how to restore faded film, I am trying to learn that now as here in Australia I have an archive with a lot of Magenta Eastman stock that I would like to preserve.

    I have found the work of Rudolf Gschwind really helpful in understanding the mathematics behind the fading, and looking for solutions, but would like to see how various people tackle these problems.
    Some of Rudolf's work is here
    Check out the movie

    and a great paper on it here:

    It really helped me understand what is happening in each channel as a result of the fading, and I have been looking to apply a mathematical solution.
  18. Wow, I did this everyday for years on far too many 1950s Fox Deluxe titles... o_O

    Custom curves, some secondaries, and pounding with a hammer during bad opticals usually did the trick for me. There's only so much you can do with the density of the blacks, depending on how the film was scanned. Multiple nodes help a lot with problematic material, too. Log tools can come in handy with some kinds of material, followed by normal primaries. Flare is also an issue, and sometimes you can even this out somewhat with power windows.

    The real trick is with density flicker and color shifts, and no conventional tools will help that. Some of the Nucoda tools will help, as will the restoration tools from companies like MTI, Pixel Farm, and Diament. Often you wind up having to use combinations of tools to fix certain problems. If the film is warped and has focus problems, good luck. I don't know of much that can fix that, but I have seen wet gate scans do miracles in certain cases. There are very advanced proprietary restoration processes that can completely eliminate flicker and shading issues; Lowry Digital/Reliance in Burbank does this better than just about anybody. Lowry also has methods for recreating (or fixing) the damaged color records (in this case, Yellow and Cyan), using the Magenta as a guide. I've seen them almost recreate good color out of thin air, even when the 3-strips were very, very damaged.

    I'm in favor of doing a light NR/enhance pass at the very end of the transfer, but this is very much an "it depends" situation where sometimes experimentation and tests are in order.
  19. Nice papers Peter, thanks for sharing. Because I didnt do lot of restoration I did it by eye. It was one year ago so I have to look again. But as Marc sad density and shifts are most important part, I think so. You can also blend chroma channel with luma, prepare it separately. I have to try it once again and will send project file.
  20. Here's a scene from a late 1950s motion picture that I color restored.
    35mm > 2K DPX


    Normally, the classic films that I see are pretty beaten up.
    As the links in Peter's post will attribute time is not kind to film for a variety of reasons.

    I have color restored over 200 classic motion pictures to date, and picked up quite a number of observations.
    Typically, I have noticed that the opticals change at a much different rate than the body of the film.
    The opticals are the opening title sequence, fades, cross dissolves, and closing credits.
    It's as though I am working on two completely different films: the opticals and the body of the film present different challenges.

    And then there is another problem, which brings us to the magenta photo in this post.
    This is a scene from inside the body of the film.
    I suspect that during this film's life someone spliced additional footage in to replace an aging or broken section of film.
    Whatever the case, it didn't match the scene before or after.

    Splitting the R, G, B channels and then using the RGB Mixer helped solve the problem.
    I exported the final grade as a LUT, which I needed to add elsewhere in the film to fix similar problems.

    I hope this helps.

    best, Jim
    Jason Myres likes this.

Share This Page