I recently posted my personally-developed method for converting camera-digitised colour film negatives to positives using GIMP and RawTherapee, but this was an overview and largely reliant on an off-site YouTube video I created. I felt a more-thorough and descriptive tutorial was warranted - hosted entirely on PentaxForums, without reliance on external video (especially given my poor video production skills :p)... so here it is! :)
-----
Colour-Accurate Conversion of Camera-Digitised Film Negatives using GIMP 2.10 and RawTherapee 5.8
There are numerous software methods available to assist with conversion of camera-digitised colour negatives - commercial plug-ins for Lightroom and Photoshop such as Negative Lab Pro, Negmaster and ColorPerfect / ColorNeg, as well as integrated tools in open source raw converters and image editors like Darktable, RawTherapee and DigiKam. I'm sure there are others I've not listed or discovered yet. Then, of course, it's possible to manually invert negatives in editors such as Photoshop and GIMP, and fine-tune colours using curves, colour balance and other adjustments. There are plenty of tutorials on the web for this, employing a variety of different methods.
Of these, I've personally tried RawTherapee's "Film Negative", Darktable's "Negadoctor", and DigiKam's "Color Negative" tools. I've also used the commercial Negative Lab Pro plug-in for Lightroom. Finally, I've carried out manual conversions in GIMP.
RawTherapee, Darktable and DigiKam's negative tools can work well, but the results are inconsistent and highly dependent on the elements within the captured negative. Colours will sometimes look OK, other times not, and in many instances you need to find neutral tones within an image to help the tools adjust for white balance and colour accuracy - the problem being, such neutral tones don't occur in every photo.
Negative Lab Pro is an excellent tool, and usually results in good-looking colours and tones... but it's not perfect. Skies will sometimes look unnaturally turquoise instead of blue, reds may look over-saturated, and shadows can have an occasional colour cast that needs to be dealt with after conversion. My biggest bug-bear with NLP is that some of its inner workings and adjustments are hidden behind poorly-explained presets that don't work consistently for every photo, or even every photo from the same roll - so you may have to go through them one by one to find the best for each individual photo. Complaints aside, NLP is an excellent tool, and it's very quick to use - so for those who want good results fast and with a minimum of effort, I recommend it. Of course, you need to own a copy of Lightroom or subscribe to one of Adobe's plans in order to use it, as it's a plug-in - and it's a commercial product, priced at USD $99 (at the time of writing).
Manual conversion in image editors such as GIMP and Photoshop works fine for inversions, but removal of the inevitable colour cast from the film base is always problematic. Curves adjustment of the R, G and B channels is perhaps the most common approach, and it can work reasonably well, but the results never seem to be spot-on - and the entire "colour character" of the film can be lost with such adjustments, so even if a nice-looking image is produced, it may not accurately represent the film stock used.
For all the difficulties with manual conversion, I very much like the complete control it provides. I also like the fact that GIMP - an open source application - is available for Windows and Linux, since I run Windows 11, Windows 10 and Ubuntu 20.04 LTS environments and like to mirror the tools I use in all three where possible. GIMP is even available for Mac, so in the (unlikely, but not impossible) event that I should end up with one of those at some future date, I'm covered.
A few weeks ago, I decided to experiment further with GIMP, to see if I could improve upon the various well-documented manual approaches to colour negative conversion. The biggest issue with all of these is colour inaccuracy - specifically, the colour cast resulting from the film base. In negatives, this is typically a strong orange, red or magenta cast, which - after inversion of the tone curve - results in a blue, cyan or green cast in the positives.
I tried a variety of approaches to deal with the colour cast in negatives:
- Colour balance slider adjustments
- R, G and B curves adjustments
- Sampling the border colour and applying a layer mask in "Divide" mode
- Sampling the border colour and applying a layer mask in "Difference" mode
- Sampling and inverting the border colour and merging a monotone copy of the image in that colour with the original based on luminosity
- All manner of other masks in different modes
A couple of these showed some promise, but all were ultimately unsuccessful.
Having reached the limits of my skills with GIMP, I decided to look elsewhere for other methods of removing the colour cast. Since I use RawTherapee as a raw converter when loading DNG files into GIMP, I began to examine the various colour adjustment tools it provided. After a good deal of experimentation, I discovered that the Colour Toning tool - when set to "Colour correction regions" mode - works very much like coloured acetates or a dichroic head in traditional darkroom wet-printing. By sampling the border's RGB values and using the Colour Toning to equalise these (i.e. the same value for R, G and B), the colour cast is removed from the negative. The adjusted image can then be transferred to GIMP for inversion, brightness and gamma adjustments, leading to a tint-free positive that more-accurately reproduces the captured scene according to the film's own colour characteristics.
There are a few additional factors in ensuring good colour. RawTherapee must be set to use the correct colour input profile and a flat tone curve, with no clipping of out-of-gamut colours or recovery of highlights. Lastly, I find it best to add a little saturation to the negative, as the neutral colour profile and tone curve can otherwise result in slightly bland colours. This is an optional step, but recommended.
Before the conversion workflow in GIMP and RawTherapee, it's essential to digitise negatives in the correct way. When digitising:
- set the camera to DNG file format
- set custom white balance against the naked light source (with diffuser, if applicable) which is used to illuminate negatives
- frame the negative to include some of the border (the importance of this will become clear later in the article)
- use the lowest available ISO setting (but not extended) to minimise colour noise in the captures
- set the camera's exposure so that the tonal range of the exposed and unexposed film (image and border) are more-or-less central in the histogram
- use the camera's highlight alert feature to show areas of over-exposure, and adjust exposure to ensure there are no "blinkies" in the photo and border (sprocket holes, if included, can and probably will be over-exposed)
- use the same exposure settings for all frames in the same roll of film
Let's now step through an example conversion.
The negative used for this was digitised using a Pentax K-5 and D FA100/2.8 Macro WR lens, mounted to a copy stand, with an EFH Essential Film Holder and Kaiser 2453 Slimlite Plano light panel. The K-5's custom white balance was set against a shot of the Kaiser light panel prior to digitising the negative.
Colour Film Negative to Positive Conversion - EXAMPLE
1. In GIMP, I navigate to and open the raw negative to be converted:
2. GIMP automatically loads the RawTherapee plug-in and selected image file:
3. I click the Exposure tab:
4. I un-check the Clip out-of-gamut colours check-box:
If this is left checked, RawTherapee may alter the representation of colours in the file to match the output profile. Whilst this would be beneficial for normal raw image processing, here it might cause colour inaccuracies further downstream in the conversion.
5. I un-check the Highlight reconstruction check-box:
As with step 4, this ensures that RawTherapee doesn't attempt to alter the representation of captured colours. In a well-exposed negative capture, this should have no effect, but it's best to un-check it anyway.
6. I set Saturation to +10:
Setting Saturation is an optional step that's also available in Negative Lab Pro where it's called "pre-saturation" (because it occurs before the conversion to positive). I've found that a value of +10 is enough, but depending on the film stock and quality of the developed negatives, I may go a little higher (up to +15 or even +20), a little lower (perhaps +5), or leave it set to zero and tweak saturation in the final positive. It depends on the individual film... but a value of 10 seems to work well across the negatives I've processed thus far.
7. I set Tone Curve 1 & 2 to Linear / Standard:
With a Linear / Standard tone curve, RawTherapee doesn't modify the captured colours, but simply represents them as the camera saw them. In normal raw processing of digital images, this would result in very flat-looking output, but for negative conversion it's an essential step in achieving colour accuracy, prior to and post the film base colour removal.
8. I click the Colour tab:
9. I set White Balance Method to Camera:
By setting the camera's cusom white balance to the light source and the white balance method in RawTherapee to Camera, the true colour of the negative is represented. The image on screen will have a noticeable colour cast according to the film base, and will typically be orange, red or magenta in hue (though other hues are possible).
10. I set the Colour Management Input Profile to Camera Standard:
Camera Standard is a relatively simple, neutral profile that's quite consistent between different camera models. The resulting colours are an honest and generally-accurate representation of those captured by the camera. This is rarely the best profile to use in normal raw processing of digital images, but for reproduction of negatives at this stage of the workflow and ahead of further adjustments, it's ideal.
11. I view the image at 1:1 reproduction:
12. I navigate to an area of the film border:
13. I select the Lockable Colour Picker tool:
The Lockable Colour Picker allows me to select a small area of an image and see the R, G and B channel values for the averaged colour in that area while I continue to work on it. As adjustments that affect colour are made, the R, G and B values will change accordingly.
14. I place the Lockable Colour Picker on the film border:
The Picker's circle is positioned so that it doesn't cover or touch anything but the border area. The average R, G and B values for the selected area are displayed. In the next steps, I will adjust the colour of the image so that these values are close to equal.
15. I enable the Colour Toning tool:
16. I set Colour Toning Method to Colour correction regions:
This is the most effective and easiest to use tool for removal of the film base colour. As previously mentioned, it appears to work in a similar manner to colour acetates or a dichroic head in traditional darkroom wet-printing. It's a clever tool that alters colour according to the American Society of Cinematography's Color Decisions List - which is worth reading up on for those who are interested.
17. The Colour correction regions tool:
Colour adjustments are made by clicking and dragging the white dot in the tool's colour matrix. Moving the dot LEFT increases the (G)reen value and decreases the (R)ed simultaneously, while moving it RIGHT increases (R)ed and decreases (G)reen. Moving the dot UP increases (G)reen and decreases (B)lue, and moving it DOWN increases (B)lue and decreases (G)reen.
The aim here is to move the dot until the R, G and B values in the Lockable Colour Picker on the border are more-or-less equal. I like to achieve an accuracy of +/- 1% or better, but in practical terms this may not be so critical. Still, I recommend aiming for +/- 1%. Final adjustment of the dot requires small, careful movements of the mouse, as the smallest inputs can have a considerable effect on the R, G and B values. It takes a little practice and a steady hand.
Since most films will have an orange, red or magenta film base, my standard approach here is to start by moving the dot to the left until the R and G values are similar. Then, I move the dot up or down until the G and B values are similar. I then make very small adjustments up, down, left and right until the values of R, G and B closely match.
18. Matching (R)ed and (G)reen approximately:
Here, I've moved the dot to the left until the R value is 82.7% and the G value is 81.2%. That's close enough at this stage.
19. Matching (G)reen and (B)lue approximately:
Here, I've moved the dot down until the G value is 81.2% and B is 78.4%. Again, that's close enough for now.
20. Fine adjustment of R, G and B:
By making very small up, down, left and right inputs to the dot, R is 80.8% while G and B are both 80.4% - within a 0.4% tolerance. That's very close indeed, and perfectly good enough for the intended purpose.
Note how the border now appears white (or bright grey). This is what I'm looking for... However, I judge this from the RGB values, not just by sight, as our eyes can be deceiving.
21. I close RawTherapee:
The RawTherapee section of the workflow is complete, so I can close it. When I do so, the image is transferred to GIMP for further processing.
22. I keep the embedded colour profile:
It's important that the colours in the adjusted image from RawTherapee remain consistent for the remaining workflow in GIMP, so here I choose to KEEP the colour profile already embedded in the image, rather than converting to GIMP's own.
23. Negative now loaded in GIMP:
The negative image has been transferred into GIMP and I'm ready to continue my work.
24. I select Colour -> Invert:
This inverts the colours in the negative to create a positive image.
25. The inverted image:
Here's how the positive looks to begin with. At this stage, it appears flat and lifeless as a result of (a) the narrow dynamic range of the back-lit transparent negative, and (b) the linear tone curve applied earlier. When you process your own negatives, they may look either too dark or bright at this stage. That's OK, and expected.
26. I click the Rectangle Select Tool:
As its name implies, this tool allows me to select a rectangular area of the image. I'll use it to select as much of the exposed photo as possible, and a little of the darker, unexposed border.
27. I select the exposed photo and border:
See how the selected area includes most of the exposed photo, plus some of the border just to the right in this example. Within this selection, the entire tonal range of the image and film base has been captured, which is critical to the next steps.
Note that the sprocket holes and film-holder border have NOT been included in the selection. This is very important, as inclusion of these would be detrimental to the semi-automated steps which follow.
28. I select Colour -> Levels:
Levels is a powerful tool for adjusting overall white point, black point and gamma in images, as well as the same three properties for individual R, G and B channels. Behind the scenes, it actually adjusts the curves for each of the respective channels, but is much easier to use than manual curves adjustment and extremely effective.
29. I click the Auto Input Levels button:
The Auto Input Levels function is quite clever. It automatically sets the black and white points of the R, G and B channels, and discards a small percentage of outlying pixels unique to each channel. This "stretches" distribution of the tonal range from pure black, through mid-tones, to pure white, and balances the colour - a kind of white-balancing function that takes care of any remaining inaccuracy in the colour reproduction. On its own, this white-balancing would be insufficient to deal with film base colour cast, but in combination with the Colour Toning adjustments earlier, it ensures good colour accuracy in the converted image.
Since I included some of the border in my selection, the result of Auto Input Levels will be that the border is rendered pure black. This is entirely appropriate, since the border is an unexposed area of the film. As such, it perfectly represents the black point for the photo, since no tone in the photo can be darker than the unexposed film.
The brightest points in the photo are now rendered pure white. Most of the scenes we capture have a broad range of tones, from dark shadows to bright highlights. In such instances, the tonal range after Auto Input Levels is applied will be more-or-less accurate. However, if the scene contained no highlights, the lighter mid or dark tones will be rendered brighter than they really were. I'll adjust for this in later steps.
30. Auto Input Levels applied:
Here's how the photo looks after Auto Input Levels has been applied. Note that the border appears black, and the highlights - for example, the sun-lit white walls of the building - are white. This is broadly accurate, although those white areas are just a little brighter than they were in the original scene, and the image doesn't look correct yet. The tonality is all wrong.
31. I adjust Output Levels white point:
The Output Levels slider control has two moveable pointers. The one at the left of the slider adjusts the black point, or darkest point in the selected area. Moving it right will increase the brightness of the black point. The pointer at the right adjusts the white point, and moving it left will reduce the brightness of that white point. Here, I've made a small adjustment to reduce the brightness, bringing the white highlights of the building to a more realistic level.
When a scene contains no bright highlights, it will be necessary to make bigger adjustments to this white point in order for the mid and dark tones in the scene to be rendered correctly. This requires some judgement by the user.
32. I adjust Input Levels gamma:
The black point of the selected area is correctly set to the unexposed border reference, and the white point was adjusted in the previous step. I now adjust the gamma - this is effectively the curve in luminosity for the tonal range, relative to the black and white points.
The middle pointer of the Input Levels slider control adjusts gamma. If the image looks too dark and "thick" or "heavy", I drag this pointer to the left. If it looks too bright and "thin" or "washed out", I drag it to the right. Here, the image looked too bright and thin, so I dragged the middle pointer to the right until the tones looked accurate.
I aim for a balanced, realistic representation of tones in the captured scene. As with white point adjustment, this requires some judgement by the user.
33. Auto Input Levels complete - Click OK:
The colours and tonal range of the selected area look good now, so I can close the Levels tool by clicking OK. It's important to choose OK rather than Cancel, as I want GIMP to remember the Levels adjustment steps for later use.
34. I select Edit -> Undo Levels:
Why am I undoing my Levels work? Because the adjustments were only applied to a selected area of the captured image!
35. I Select All:
I select the entire image so I can re-apply the previous adjustments across this wider selection.
36. I re-open the Levels tool:
37. I select the previous Levels adjustments:
The Presets drop-down box in the Levels tool contains any named presets created by the user, but also the last-used adjustments. I want to select the most recent adjustments - as shown by the date and time stamp (in my drop-down box, it's the item at the top of the list).
38. Adjustments re-applied:
After selecting the last-used adjustments, they have been applied to the entire selected image, and I can click OK to close the Levels tool.
39. Conversion complete:
The conversion is complete and I have a positive image with accurate colours and black point, and realistic tonal range!
40. I select File -> Export As:
Having completed the conversion, I create a master TIFF file. This can be used as the starting point for any future exports in different formats, resizing, crops etc.
41. Naming the master positive file:
The master positive TIFF should be named according to the user's preferred conventions. I like to add "_positive" to the file name. The extension must be ".TIF". When the file has been named, I click Export.
42. Export the master TIFF:
Here I can choose the options for the TIFF file. I always accept the defaults, but if you have other preferences, feel free to select them as required.
And that's the end of the conversion workflow :)
Presented here, it seems like a lot of work - but that's because I've laid out every step with an accompanying screen-shot. With a little practice, it takes just a couple of minutes or so to convert a negative in this way, from opening the negative DNG to exporting the master positive TIFF. But you can make the workflow even more efficient...
After steps 1 through 10, you can save the current settings as a Processing Profile in RawTherapee. I'd recommend naming it something like "Colour Negative Conversion". For each subsequent negative that you wish to convert, simply apply this saved Processing Profile in RawTherapee and you can move straight on to the Colour Toning steps.
If, as I recommend, you use the same exposure settings to capture all the negatives from the same roll of film, there's one more level of efficiency you can bring to the workflow...
After completing the Colour Toning steps to remove the film base colour, save the Processing Profile with a name relevant to the roll of film you're processing. This profile can then be applied to every frame from that roll - meaning you won't need to repeat any of the manual RawTherapee steps for subsequent frames. Since the Colour Toning steps can be a little tricky and time-consuming, this is a huge time-saver. However, don't be tempted to use this profile for other rolls, as even for the same film stock, the properties of the film can change slightly ;)
I hope this tutorial proves useful to some of my fellow members. Please let me know what you think, especially if you've tried it for yourself or you're thinking of trying it. I'd be delighted to answer any questions to the best of my ability, or to help if you run into problems.
I will post any updates to this method in both the original post and the end of the thread, so please check back from time-to-time.
Enjoy, take care, and best wishes :)
Mike
|
