Film Negative Invert and Processing in L* Gamma FAQ

[shijan]

File opened in ACR and imported into Photoshop with no changes is a bad idea because ACR always apply additional tonal curve to image and it destroy original tonality of film source. Processing film negatives in sRGB also very bad idea because sRGB color space is way smaller than native film color space and negatives processed like this can't be counted as real digital copies. Mistakes in digital workflow and color management are not the same as variations in film color. But from some point of view it all depends of luck and presence of extreme saturated colors. Sometimes even images processed in sRGB and ACR may look nice.

 

"I still don't fully understand why the red channel has a tendency to become overly saturated by using a filter. It needs further investigation. However, I see use of a mask correction optical filter as a step forward in the workflow."

 

because you process in sRGB, because ACR add additional contrast to source image and because blue filter changes color balance too much.

Edited March 24, 2021 by dmitry_shijan

[shijan]

Ok, back to technical research.

Here are some examples of camera calibration importance. This example was taken with Fujifilm X-Trans sensor and low CRI consumer LED backlight. As you can see proper camera/scanner calibration applied to image may fix a lot of problems caused by poor quality lighting. And as expected, Lab cLUT ICC profiles produce the most accurate colors.

 

 

 

[digitaldog]

And as expected, Lab cLUT ICC profiles produce the most accurate colors.

I don't see. Got a dE report of accuracy (reference and result)?

[shijan]

Just look at Yellow patches, they are the most notisable. Also weak greens. Due low quality light source, yellows have strong green tint in image that use original camera input profile. It is partially fixed by Matrix-based ICC profile. And very way better fixed by Lab cLUT ICC profile. Edited March 24, 2021 by dmitry_shijan

[digitaldog]

Just look at Yellow patches, they are the most notisable.

That tells me nothing about color accuracy. When you tell me the Lab values of the source yellow, then the result, at least I can calculate color accuracy (of one patch). That's all I'm asking (and ideally, a dE report of ALL the colors in the target with reference Lab values).

[glen_h]

Not wanting to figure this out, I have gone for using DSLR for slides, but actual scanners

for negatives. Even though the scanner designers aren't perfect, I trust them to get it

right, more than for me to figure it out from a DSLR image.

 

I have a few film scanners, of a range of quality, speed, and cost, to use,

including a Pakon F135, faster for whole rolls of negatives.

[shijan]

digitaldog, It is subjective opinion that image looks better after calibration and i don't have any special test measurements. The difference is easy to see with naked eye, so i really never thinking about inspecting this in more scientific way.

Here is link to reference data file for this test chart scan if it helps, i guess it should have all reference L a b color target values there http://www.colorreference.de/targets/E130401.zip

 

Here are also visualizations of ICC profiles generated with this slide film TI8.7 chart compared to sRGB. As you can see original color space of the slide film is way larger than sRGB and AdobeRGB color spaces. Film negative color space arranged is near same way. This is why it is not recommended to invert and process your scans in tiny sRGB or AdobeRGB color spaces. Use ProPhotoRGB color space and keep all original colors.

[rodeo_joe1]

because you process in sRGB, because ACR add additional contrast to source image and because blue filter changes color balance too much.

Then why does it not happen to the image that was WB adjusted to neutralise the mask? The exact same procedure was applied to that.

 

The mask is there to compensate for weaknesses in the cyan and magenta dyes of the film. So excess red would seem to indicate that the optical filtering does too good a job of reinforcing the cyan layer.

 

Of course, the negative was meant to be printed on paper that also has imperfect dyes with some yellow bleed into the magenta and cyan dye layers, which our RGB digitisation doesn't suffer from. Therefore the legacy colour paper would automatically introduce some hue rotation and attenuated reds and greens. As well as a very steep 'gamma' curve.

 

Maybe the more one attemps to reproduce the original chemical process, the more kludgey one has to make the digital process. So I don't see a hue rotation as any form of 'cheating'. Just part of whatever it takes to get to an acceptable end result.

 

I'm not sure where those film colour spaces shown above come from, but they look mighty suspect to me. How can CMY dyes form a near RGB triangle? The dyestuffs have to combine to form additive primaries, and in doing so they automatically lose luminosity. Therefore very saturated reds, greens and blues can only occur at low brightness levels.

Edited March 25, 2021 by rodeo_joe|1

[Ed_Ingold]

It was inevitable that someone would come up with a nearly foolproof solution (ignoring the ingenuity of fools) to color negative conversion. www.negativelabpro.com has a solution which works in conjunction with Adobe Lightroom as a plugin, which fits the need nearly perfectly. There is a fully functional trial version, good for 12 conversions, but as many views as you wish.

 

This example is shown as converted, without further adjustments. It is from a 6x6 medium format negative, which I didn't even crop properly (very hard to see the borders), "scanned" with a Sony A7Riii and Sony 90/2.8 Macro lens, from a RAW file. In my experience, it is best to save the raw scans, which makes them available for re-conversion at a later date. This is one example, but after a brief trial, I bought the software and got similar results on a couple dozen other conversions.

 

I am especially pleased with the rendering of the sky, which often comes out green or magenta using other conversion schemes. The software will do batch conversions too, which can be synchronized in Lightroom.

 

Edited March 25, 2021 by Ed_Ingold

[shijan]

Yes, there are a lot of other tools. I collelct some very subjective personal thoughts about alternate apps and tools for negative film processing that i use and don't use:

 

darktable negadoctor standalone app

darktable - the photo workflow software

Free all-in-one tool. Allow native RAW workflow. Nice debayer quality and nice options for Fujifilm X-Trans sensors. For my personal taste it is too "consumer" and too abstract to use. It works OK with simple images, but don't allow to fine tune films with complicated colors and unusual tonality. darktable allow to use custom working color spaces but looks like operates with film negatives only in sRGB gamma. Controls may feels not too intuitive for the first look, but there are a lot of video tutorials on Youtube and Vimeo.

 

Negative Lab Pro plug-in for Adobe Lightroom

Negative Lab Pro

Overall it produce same result as darktable negadoctor. Same time it feels like slower and less comfortable to use. It have some amount of useless consumer presets and "looks" variations.

 

ColorPerfect plug-in for Adobe Photoshop

ColorPerfect Plugin for Adobe Photoshop | PSE | PL

Worst investment ever. Overpriced app that seems like designed to made impression of something "very complicated" and "very important". Worst ever made Interface. Very slow processing. Low quality results. Clipped dynamic range. Hard clipped contrast adjustment. No real color management. Can't read camera input color profile. Communication with developer is possible but goes to nothing. Instead of clean answers to simple straight questions he wrote back giant pseudo technical texts that explains nothing and suggest to process images in sRGB color space.

 

FilmLab standalone app

FilmLab

I check demo version and it feels like some joke. Very limited controls. Colors are always off and it feels like it attempt to prettify image and add some sort of fake "filmic look" LUT on top.

 

Grain2Pixel plug-in for Adobe Photoshop

Grain2Pixel – FIlm Photography Tools

I attempt to test Grain2Pixel, but can't run it because it require newest version of Photoshop. Very strange limitation.

 

CNMY plug-in for Adobe Photoshop

CNMY film inversion

I didn't test this app yet, but hope to check it someday.

 

Negmaster plug-in for Adobe Photoshop

NEGMASTER - A brilliant Photoshop plugin for converting negative and slide film

I didn't test this app yet, but hope to check it someday.

 

SilverFast NegaFix

I didn't test this app yet, but hope to check it someday. From online video examples tutorials user interface looks very outdated, ugly and not logical.

 

Analogue Toolbox for Capture One plug-in for Capture One

Analogue Toolbox for Capture One - Curious Photography by Michael Wilmes

I didn't test this app yet, but hope to check it someday.

 

SILKYPIX Negative film inversion tool standalone app

[Pro10] The Negative film inversion tool | SILKYPIX

I didn't test this app yet, but inversion examples on their website looks very bad. Hope to check it someday.

 

VueScan standalone app

VueScan Scanner Software for macOS Big Sur, Windows 10, and Linux

This is another example how to make simple things in too complicated and non human friendly way. It works, it scans, but i hate the logic of this app and hope i will never to use it. Also it may produce strange pixel patterns artifacts with some film scanners models. Have no idea was this problem fixed or not, but i always recommend to use original scanners drivers and original software if possible.

 

PhotoLine standalone app

PhotoLine: Photo Editing, Vector Editor, Design Software

Sort of "Secret weapon". Without digging into details, from technical point of view compare to Photoshop it is like intergalactic spaceship compare to basic airplane from 1970 era.

PhotoLine was started in early 90's (near same time as Photoshop). It is made by very small team of developers in Germany and not too advertised as most other graphic editors. Developers focused mostly on technical side and don't care too much about interface look and pretty icons. So visually it is just not as pretty looking as Photoshop and i guess this is the main reason why many people reject to use it. Tools and logic in PhotoLine in most cases are superior to Photoshop, but some tools in some places feels slightly limited and require some updates and improvements.

I started to use it about year ago and same time start to help on forum with bugfixes and ideas as a volunteer. During this time developers add huge amount of changes, bugfixes and improvements. I also start PhotoLine UI Icons Customization Project to make it look more user friendly PhotoLine UI Icons Customization Project by shijan on DeviantArt

 

Iridient Developer standalone app

Iridient Digital

Very solid, powerful and simple RAW editor. Great debayer options and sharpen quality. Probably the best one for Fujifilm X-Trans sensors. A lot of other unique options. Quick and easy to use. UI is very basic and not too pretty.

[shijan]

There are a lot of apps with some abstract sliders that may invert negative. Some do it well, some don't. The main point of this thread was to show the core of the invert and RGB Levels process, to help to understand hidden technical problems, to explain how tonal arrangement works in L* gamma and how camera calibration affect colors in final image...

 

By the way, Ed_Ingold, here is image converted from your JPEG example with method explained first post :) Jpeg only was converted to ProPhotoRGB and L* gamma for start. I done this just as extreme example. I don't suggest to process film negatives from JPEG sources.

Edited March 25, 2021 by dmitry_shijan

[digitaldog]

digitaldog, It is subjective opinion that image looks better after calibration and i don't have any special test measurements.

Then you can't speak of color accuracy. You can speak of pleasing subjective color.

I can speak of accurate color when I have Lab values for reference (what the color should be, and you have that with such a target) and resulting color values:

Your gamut maps, which are not ideally constructed*, tell us nothing about color accuracy.

* ColorWiki - Color Management Myths 26-28

If you wish to speak of color accuracy, you need to understand basic colorimetry.

[shijan]

The main problem with film negatives that they are not accurate by nature. Even same roll many variations. it is impossible to print tech chart for calibration on them natively. So the closest compromise is calibration test chart printed on slide film.

It also not clinically perfect because each slide film brand also have some own character. All i done is just use IT8.7 target and provided data for it from Affordable IT 8.7 (ISO 12641) Scanner Color Calibration Targets and build input ICC profile. This works subjectively well with film negatives from scanners as well as from DSLR cameras, this fix problems with low quality light sources and makes final image look subjectively better.

 

By the way, digitaldog, i watched your videos long time ago and learned a lot with them. They really inspire people to digg into color management stuff.

Edited March 25, 2021 by dmitry_shijan

[rodeo_joe1]

Well, that wasn't too difficult.

I just aligned the end-points of the RGB histograms using the editor on my phone.

 

After inversion, the red histogram was nearly disappearing off the LH edge, and the blue was cramped up against the RH side. Which is why getting rid of at least some of the orange mask is a good idea.

[shijan]

The only problem that sunset became yellow-green in your version instead of orange:)

 

Formally the whole concept of "Remove Orange Mask and all colors magically became correct" is incorrect from some point of view. Orange mask is not a some separate evil part of the of image, it is a part of the film negative and it is linked to relations between colors, so is not something that should be somehow removed or just hard clipped to white. It is a symmetrical system that was build for film negative + print paper.

Technically In 16 bit it makes no difference for Photoshop or any other app if orange mask "removed" or not. It is all just a adjustment (alignment) of dark and light points in channels. The main problem is how to keep relations between these aligned channels as original as possible. So when some people first attempt to "remove orange mask" with aligning RGB channels by picking WB, and then "fix colors" by AutoLevels, they formally in best case scenario just move same data from one place to another and back. In worst case scenario these additional "movements" generate shifts between colors and as result "corrected" image require more additional corerctioin to adjust colors back to subjective normal look.

[rodeo_joe1]

The only problem that sunset became yellow-green in your version instead of orange:)

Yeah, well I only spent about a minute on it. With a very limited editor and no reference image for it.

 

Besides, how do we know what the 'correct' colour should have been? Because that sun looks very pale for a sunset scene.

Formally the whole concept of "Remove Orange Mask and all colors magically became correct" is incorrect from some point of view.

From your point of view.

'All colours becoming magically correct' has never been my argument. What optical neutralisation of the mask does is to align the contrast and brightness values of the cyan, magenta and yellow dye images in the film. Otherwise the histograms are offset and of different lengths after digitisation. Equalising them means you have more digital bit depth, and more equal RGB images to manipulate. And they all lie on the same part of whatever gamma curve was applied to the RAW file.

 

It's simple optics. Just look at the curves for some colour negative films. The 3 curves are vertically separated due to the mask, and this translates to a horizontal separation of histograms after digitisation - unless the mask is neutralised.

Edited March 25, 2021 by rodeo_joe|1

[shijan]

Few more very basic extreme examples:

 

8 bit Image processed in sRGB gamma and sRGB color space.

AutoLevels Clip was set to 0.05 %

As you may see from RGB histogram in Levels tool, relations between channels became very non uniform after invert. Red channel stretched in very strange way. You can see that greenish color cast there from start.

Tonal arrangement require large amount of additional manual adjustments to bring back things to normal. All these problems caused by too tiny for such editing sRGB color space and not too "symmetrical" sRGB gamma.

 

 

Same 8 bit Image converted (transformed) from sRGB to ProPhotoRGB color space and L* gamma.

AutoLevels Clip was set to same 0.05 %

Due larger color space and due more "symmetrical" L* gamma, there is enough of empty space for adjustments in each channel. Relations between channels stay uniform after invert. Tonal arrangement looks nice and require very small amount of adjustment.

 

[digitaldog]

(ISO 12641) Scanner Color Calibration Targets and build input ICC profile.

That's the reference indeed. But without the resulting color values, it is like one hand clapping. One cannot calculate color accuracy with just the references.

[Ed_Ingold]

There's more than one way to skin a cat, but I am done fiddling. I have hundreds of rolls which need scanning/conversion and life is too short to twiddle Curves and Levels. The photo above is pretty simple, since nearly everything is black except the sunset. It was taken in 2004, just before a hurricane devastated Ft. Meyers, FLA. The last roll of film I took was in 2014, in a Leica M3 for comparison with a Leica M4P digital camera - same place, same subjects, same lens and same time. I should have made "last" bold. After this comparison, my M3 retired.

[Ed_Ingold]

Unfortunately, you cannot create a reliable profile for shooting negative film. However you can create a profile from a color chart which can be implemented in post. I've gone through several iterations of software which can create such a profile, but currently use XRite software with good results. Consistency is contingent on using the same light source and exposure. Ambient light at the time of capture also affects the results, e.g. under a forest canopy or open shade, exceeding the limits even a profile can correct.

 

I create a profile, Assign it to the image as the color space, then Convert the color space to a standard version, e.g., AdobeRGB. Silverfast will create and use profiles for scanning positive film, but not negatives.

[Ed_Ingold]

My brother came up with a pithy phrase which fits some of the responses. "When you value your time at $0, a $99 plugin has an infinite cost/benefit ratio." Edited March 25, 2021 by Ed_Ingold

[glen_h]

The orange mask comes from colored couplers that are supposed to have just the absorption that the actual dyes should not have.

 

That way, the difference between coupler and dye is exactly what it is supposed to have.

 

In addition, color negatives have a lower gamma than black and white negatives, which makes print exposure harder.

Pretty much you need an enlarging light meter to print them. The rule for filter selection is half the complementary

color of the filter that looking through makes the color look right.

 

That makes scanning also twice as hard to get right, as it will be twice as sensitive to any color shift.

[rodeo_joe1]

Here are two more camera 'scans'. One with the mask base colour filtered out, and another with the camera WB set to that of the flash light-source (+ opal diffuser). Below each is their set of histograms.

 

Non mask-corrected version:

Whoops! The red histogram appears to be crushed up against the RH edge of the curves graph, despite no sign of overexposure in the camera.

 

Now with elimination of the mask:

All the histograms are now complete, and sit in line with each other.

 

Let's see how that affects the inverted result.

Mask-biased inversion:

 

And mask-eliminated inversion:

 

Is it not obvious which version will be easier to adjust to acceptable colour?

In fact, just using Auto-color Adjust in Irfanview got me this -

Not perfect, but for a single-click solution it's pretty good. And the same procedure on the over-blue masked version was nowhere near useable.

 

A bit more work got me this result.

The colour isn't checker-card accurate, but the area is called 'Golden Cap' and I wanted to emphasise the yellow tones of the cliff against the blue sky.

 

That was taken on Kodak film. Here's another on Fuji stock, using the same mask elimination procedure.

The Fuji film has much less tendency to exaggerated reds, but I'm not liking how it handles highlights.

 

Trying to iron out all the quirks of various negative films is a pretty fruitless exercise IMO. And a pleasing result is all I'm after.

 

BTW. I experimented both with MakeTiff and applying a Prophoto profile to the RAW import. MakeTiff worked for some negatives - on the same roll - and failed spectacularly with others. The Prophoto import tamed Kodak's oversaturated reds, but then over-emphasised greens. So overall not much of a time saver.

 

FWIW. The contrast mask has done its work once the film is developed, and filtering its base colour away doesn't affect the relationship between the Cyan, Yellow and Magenta images in the film layers. Because that's exactly what happens in the opto-chemical process it was designed for.

So, no; the mask is not 'evil'. Just a damned nuisance!:eek:

Edited March 27, 2021 by rodeo_joe|1

[shijan]

Not again please :)

 

First of all your Non mask-corrected version source is clipped in red channel, so you just operate with already broken image from start. Delete it and re-scan without clipping.

 

"experimented both with MakeTiff and applying a Prophoto profile to the RAW import"

This is that you never should do, because you just destroy color management when apply ProPhotoRGB (or any other working ICC profile) instead of proper input camera profile. This is main problem of ColorPerfect/MakeTiff developer - he did not provide proper color transformations system form camera input profile to working profile.

At the first step image always should be CONVERTED from camera/scanner input profile to working profile.

 

"Is it not obvious which version will be easier to adjust to acceptable colour?"

No it is not. And i post extreme examples and link to explanation earlier.

 

 

All i can suggest - read faq in first post and learn how color management works. Until that i can't see any point to explain you any things further.

 

Debayer RAW file and convert it from Camera input ICC profile to ProPhotoRGB with L* gamma custom-made ICC profile -> Invert -> Apply RGB AutoLevels -> Recover back clipped data from RGB AutoLevels -> Adjust tonal curve and Contrast.

 

Compare test results to your methods. Maybe later you will see your mistakes. I am not interested anymore to discuss with you problems and mistakes that i already fixed for myself 5-7years ago.

[shijan]

P.S. to everyone:

I want to notice again - this thread is not about some abstract workflow and some guesses how to process film negatives. It is about technically correct, well established and polished workflow that provides quick, solid and predictable final results as described in first posts.

So please make sure you at least read the topic and understand how it works before post something here.

If you want to discuss, confirm or deny something do some compare tests with your own workflow and workflow described at the start of this topic, provide decent info of your workflow and post source RAW test images.