[Play_RAW] Scans from old negative film > heavy color cast

Hope I am not off-topic here. It’s not really a Raw Play thing, it’s more about saving old negatives film.

Film is about 20 years old. Somehow it lost lots of blue. I scanned it with a Nikon Coolscan without any color correction. This is what it looks about:

Here a link the 8bit TIF (~65Mb)


What I did in Darktable was using Tone Curve > Lab Channels > Set the white point in the Lab B (the surfboard) to the center line, and kept a straight line.

That makes the image look a bit more normal. Still, no blue sky.

What I look for is basically to get ‘okay’ with a few clicks. I am aware that one can do much more, but I have a about 1000 images to scan and my aim is not perfection here. A believe the quality of the negatives has quite deteriorated over the the last 20 years, so my expectations are not that high. And some are way worse than this one.

1 Like

Nice image. Please license your image; see: PlayRaw stuff to keep in mind.


Hi @st.raw,

you can use color balance module to remove color cast quickly. As in your example above, you can use the surfboard surface as a reference. Choose color picker for hue from highlights : gain/slope and mark an area on the surfboard. Result:

Additionally you can use the color correction module to make further corrections. To my taste, the warmer dark tones look most pleasant for this picture:

By the way, I don’t think the sky in this picture is blue. To me it seems to be rather cloudy. This can be also seen from the fact that there are no strong shadows in the rest of the picture.

1 Like


Image11_01.tif.xmp (12.8 KB)


@yteaot Nice. What is your general methodology here, as I am away from the computer.

@st.raw Old negative film only kept its color for a few years. New film colors will last for decades. Magenta fades the slowest. The compliment of magenta is green. So your positive result will be very green. And it is. The cyan and yellow fade faster than magenta and at different rates. The result is a major color imbalance. I’ve mostly scanned Kodachrome slides and have seen no color degradations after 40+ years, so I’m relatively inexperienced at the color rebalance methods.

Image11.tif.pp3 (13.0 KB)
Image11-1.tif.pp3 (14.0 KB)


-First Colorbalance / neutralize colors.
Then Color Zones / almost all the colors off but keeping some red
-Then repaint with color look up table + mask. Separately green forest, bluish water and sky
-Repair with retouch edge on left bottom and white dock
-Some adjustments with filmic and crop and rotate

I don’t know what the original colors were like. I chose the colors that seemed to fit the picture’s condition. So I did not try to restore colors, I made new ones


second version with less magenta skin tones, mainly rgb curves and channel mixer :slight_smile:
Cropped with a 16:10 ratio

Image11.tif.pp3 (13.3 KB)


@yteaot, really impreessive what you did!

I just tried to do some general corrections.

Image11.tif.xmp (4,0 KB)

Hello @st.raw. If I undestand correctly, you need a quick solution (a click or two long) for restoring your slides. Since the original color seem to have degraded considerably, may I suggest that you keep them in B&W form and the ones that are really special can be looked into restoring in color. Here is my three click solution using RawTherapee 5.5.
1.Turn on the Black and White module
2. Turn on Local contrast module.
3. Turn on Haze Removal module.
You get this:

1 Like

This was fun, thanks for sharing! Here’s my take…


Thanks for all replies. Interesting to see the different approaches. My comments:

20 Second fix, but there is more in it.

I need to look into that. Interesting approach.

Saturations seems a bit low but I really like the tonality. Quite a lot of steps though. Need to study that further too.

That scan isn’t that bad. I have worse, much worse.

Similar colors to @yteaot, which I really like a lot. I don’t really use Rawtherapee, had to look up what pp3 is associated with. Can’t saw much yet, but the menu looks much better organized. I love that one can see what module is on/off instantly. Update: Tried your method with Darktables Channel Mixer - so far failed. DT’s Channel Mixer is really user unfriendly compare to RT. Very confusing and easy to delete what you’ve done when you switch off a channel (will instantly disappear from the pipe and not come back) - One more try: Using the Channel Mixer for each color individually seems impossible.

You sort of set a negative curve in Lab B. Interesting. Very effective.

I love b&w, mainly for artistic reasons. My scan doesn’t really qualify for that. As long as I can try to squeeze some color out of it I will try that.

Very impressive. Can you post the side car file please?

@st.raw I generally like @age’s handle on colour. Getting skin right is pretty important.

I don’t have a colour managed workflow and my screen is old and low quality, so I am not confident in processing the image. :blush:

1 Like

I’m afraid I don’t have one that can be useful, sorry. however, the bulk of the work is done with a channel mixer, which is available in many programs. that will get you 70% there. the rest are more local tweaks and a global increase in contrast and saturation. if you want something quick that can be applied in batch, I think that channel mixer alone would give some decent starting point

One click …Photostylizer filter in Google’s free Color Effects module of NIK…could be tweaked and saved as a recipe for your batch edit

I think the options were Varitone and applied at maybe 20ish percent

Thank you for the play.
Edited with GIMP and G’MIC.

First image scaled down to 25% for color work, for using an old laptop, scaling down images means less time for processing.

The following filters are used:
“Tune HSV Colors”,“Mixer” [PCA]", “Transfer Colors [Histogram] and [PCA]” (original with modified original back and forth, like a loop).
“Local Processing” (only for colors, thereafter “Transfer Colors” with edited and/or original, then mix result with layers).
“CLUT from After - Before Layers”.

Then on original scale:
“Apply External CLUT”.
Again a little correction with “Curves” and a little bit “DCP Dehaze”.
No masking, retouching, sharpening or noise reduction. Only global filtering.

Maybe the result isn’t spectacular, but I had fun and learned something:

1 Like

Looking at the histogram I’m getting of the ‘image11.tif’ file, I’m amazed by what people are able to do with it.

I’m reading it as an 8bit file, with like no blue in it at all. I don’t know if it’s a positive scan (slide) or a negative scan (c41/cross) but normally you would fix a huge balance problem like this during the scan.

If you want to fix it afterwards, at least save it as a 16bit file so you save the dynamic range a bit of channels that are too low (like blue here).

The best tricks that seem to work are things that ‘rotate the hue’ of the entire picture, or ‘colorize’. Which means they’re sort of reconstructing a new blue channel :).

1 Like

The image in this topic can’t be corrected with only channel mixing and colour balance, you have to retouch it like @yteaot did.

Not only the blue disappeared, but after channel mixing and colour balancing it looks like the green colour is now more blue then what is supposed blue. Like green and blue swapped in the spectrum. I couldn’t switch these colours back with the knowledge and tools that I have (rotating hue for example), without creating large artifacts.

But when trying to “swap” green and blue/cyan, I got the impression that the shorts of the man in the rowboat were more blue than green (in the original colours of the photo just taken). Difficult to find out if that idea is correct.

cheating a bit by not using Darktable… but I also still think you need a 16bit version to save whatever data you can from the blue channel to get a good result.

I tried to merge the green + red channels to ‘fabricate’ a fake blue channel. I started by doing a 50/50 for red/green but that quickly was way to red. In the end I used mostly green and just a bit of red to make a new blue channel.

In the end I ‘color balanced’ it by using a levels-(gamma)-adjustment to align the ‘median’ values of the individual channels with the median of the whole image.

I did it in Affinity Photo, but Photoshop or other tools with layers could work just as well I guess.

Really, for the (probably not) last time: scanned pictures need 16 bit for the color-correcting afterwards.

Even if we have a blue channel that is 0.5% of the exposure of the red + green channels, that still gives us around 328 values to work with, which should work out fine to make a final 8-bit 0-255 image.