Narrowband RGB Film Scanning

For a long time, the standard recommendation for camera-scanning colour negative film was relatively simple: use a good, high-CRI broadband white light source, photograph the negative in RAW, and deal with the colour during the inversion process. That approach works, and very good results can be obtained from it. But there is another way of approaching the problem. Instead of giving the camera the entire visible spectrum and asking it to disentangle the colour information afterwards, we can control the spectrum of the light used to scan the film in the first place. This is the idea behind RGB film scanning and, more specifically, narrowband RGB scanning.

The basic idea is surprisingly simple:
Instead of illuminating the film with white light, illuminate it with separate, relatively narrow red, green and blue bands of light.

To understand why this can be useful, it helps to first understand what a colour negative actually contains and what happens when a digital camera looks at it.

Colour negative film and the three dye layers

A colour negative contains three image-forming dye layers: cyan, magenta and yellow. These dyes are complementary to red, green and blue light:

• Cyan primarily absorbs red light.
• Magenta primarily absorbs green light.
• Yellow primarily absorbs blue light.

In an idealised world, each dye would absorb only its corresponding colour and transmit everything else.

Real photographic dyes are not ideal filters, however. Their absorption spectra are broad and overlap. A cyan dye, for example, does not absorb only red wavelengths. Some absorption extends into neighbouring parts of the spectrum.

This overlap is part of the reason that colour-negative inversion is a more complicated problem than simply turning the negative upside down and inverting its colours.

There is another complication: the digital camera.

The camera sees colour differently from the film

Most digital cameras use a Bayer colour filter array over a monochrome image sensor.

The individual photosites are not inherently red, green or blue. Instead, tiny colour filters are placed over them in a pattern, usually containing twice as many green-filtered photosites as red or blue. To output an image in colour the camera uses a debayering algorithm to generate a colour image which is an coloured approximation of the scene in front of the camera. A pixel covered by a red filter does not respond exclusively to red wavelengths. It has a spectral response extending over a range of wavelengths. The same is true for the green and blue filters. As a result, the three camera channels overlap. This means that when a colour negative is illuminated with broadband white light, the red, green and blue channels recorded by the camera contain mixtures of information from the different dye layers. This combined with the spectral overlap between dye layers in the colour negative is usually referred as crosstalk

Here is a simplified example: Imagine illuminating a negative with red light. Ideally, we would like the camera to record:

Red channel = signal
Green channel = 0
Blue channel = 0

But a Bayer sensor might instead produce something more like:

Red = 100
Green = 20
Blue = 5

The green and blue values are unwanted information. They are the result of the spectral response of the camera's colour filters. The exact amount of crosstalk varies considerably between cameras. This becomes particularly important when trying to recover accurate colour from colour-negative film, because the signal we are trying to measure is already a mixture of overlapping dye responses.

The orange mask

Colour negative film also has its characteristic orange mask. The mask is not simply a colour cast that was accidentally introduced into the film. It is part of the design of colour-negative film and is related to the unwanted absorption characteristics of the image-forming dyes.

The colour-negative film system was designed as part of a larger colour-printing process. When a negative is printed onto colour photographic paper, the spectral characteristics of the negative, enlarger light and photographic paper work together as a colour system. RA-4 paper has its own spectral sensitivity and the enlarger provides controlled colour illumination, traditionally through colour filters or specialised light sources. The important point is that a traditional colour print does not work by asking a generic camera to measure the negative's colour. The entire optical and photographic system was designed around the spectral characteristics of the materials. When we replace that system with a digital camera, we have to recreate some of this colour separation computationally or optically.

The darkroom connection

The idea of separating colour information with light is therefore not new. Colour enlargers have historically used separate colour channels or carefully controlled spectral illumination when exposing colour paper. Some enlargers use filters to modify a broadband light source. Others use more specialised narrowband light sources.

The Philips PCS 2000 enlarger, which came out in 1981, is an interesting example of a system using controlled spectral illumination.

Here is an excerpt from the introduction out of that same manual:

The ETC System uses the three primary colours for additive printing - hence Tri, but with one exposure time therefore TRI-ONE. Instead of unstable filters of varying densities, the Tri-one system uses permanent, high quality filters with narrow-band colour channels closely related to the colour sensitivity of photographic paper. The 'amount' of the light and hence colour 'density' is electronically controlled by precisely adjusting three voltage-stabilized lamps in the unique Philips ETC light source. The control of the light source and therefore of colour balance is carried out using the separate Electronic Tri-one Colour Unit which displays in a logical and straightforward way all colour responses.

The principle is the same one we are interested in today: control which wavelengths reach the photographic material. In a darkroom, the colour paper provides the spectral response that determines how the different parts of the negative are interpreted. With a digital camera, the sensor provides the response instead. That difference is crucial.

Solving Colour Separation

There are broadly two places where we can deal with the colour separation problem: physically, during capture or digitally, during processing.

A conventional white-light camera scan puts most of the burden on the digital side. RGB scanning moves more of the separation into the capture process. This does not mean that software becomes unnecessary. Rather, the software starts with a cleaner and more controlled set of measurements.

Specialised film scanners have been using separate RGB illumination for many years. A common approach is to illuminate the film sequentially with red, green and blue light and capture each colour with a monochrome sensor. Because a monochrome sensor has no Bayer colour filter array, it does not have the same RGB channel overlap as a normal consumer camera.

The scanner can therefore do something very direct:

red light → red measurement
green light → green measurement
blue light → blue measurement

The three measurements are then combined to produce the colour image. This principle can be found in professional film-scanning equipment and is one of the reasons that RGB illumination is not a new invention of the DIY scanning community. The challenge for camera scanning is that most of us are using a conventional colour camera rather than a dedicated monochrome scanning sensor.

Some of the most advanced scanning systems for scanning and restoring cinema film make use of this principle. Some scanners include the Arri Arriscan and Scannity which criterion collection uses restore their film. In the promo video below for the Arriscan one can see the R G B light going of separatily while scanning the film

ARRISCAN Commercial RGB light working

Photographic film scanners from the 90s and early 2000s also made use of RGB light to achive colour separation from colour negatives. For example the Nikon Coolscan lineup and the lab scanner Fujfilm Frontier SP-3000.

RGB filters

One way to create separate RGB illumination is to start with a broadband white light source and filter it. A high-CRI white LED contains a broad range of visible wavelengths. We can therefore place a red, green or blue filter in front of it to select the part of the spectrum we want. There are several ways to do this.

A straightforward approach is to use three separate filters:

• red filter
• green filter
• blue filter

The film is photographed three times, once through each filter. This gives us three separate exposures.

Another possibility is a tri-band filter. Instead of having a filter that passes one continuous region of the spectrum, a tri-band filter is designed to transmit three relatively narrow spectral regions corresponding to red, green and blue. This allows all three bands to reach the film simultaneously. The advantage is that we can produce a trichromatic light source from a broadband white source without needing three separate light sources. The disadvantage is that we are still starting with a broadband source and then throwing most of its light away through filtration. The exact transmission curves also depend on the particular filter.

From filtered white light to RGB LEDs

A more direct solution is to generate the narrow spectral bands at the source. Like seen before in the enlargers and film scanners.

Instead of:
white LED → filter → RGB light

we can use:
red LED + green LED + blue LED → diffuser → RGB light

This is the principle behind narrowband RGB scanning lights. The individual LEDs can be selected for their spectral output, allowing the scanning light to be designed specifically around the spectral behaviour of film and the camera sensor. This is where the modern DIY RGB scanning movement becomes particularly interesting. Not all RGB LEDs are equal. A general-purpose RGB LED designed for lighting or display applications may use relatively broad or less-than-ideal wavelengths for film scanning.

For camera scanning, deeper red and deeper blue wavelengths can be advantageous because of the spectral sensitivity of typical Bayer sensors. Jack Whittaker's (jackw01) Scanlight project is one of the best-known open-source implementations of this idea.

The current Scanlight design uses approximately:

• Red: 665 nm
• Green: 525 nm
• Blue: 455 nm

These wavelengths were chosen with the spectral response of typical digital camera sensors and the film dyes in mind. Scanlight is an excellent resource if you want to explore the hardware in more detail.

The choice of wavelength matters because the Bayer filters in a camera sensor are not equally sensitive to every part of their nominal colour range.

Jack Whittaker's measurements and experiments indicate that wavelengths above roughly 650 nm for red and below roughly 450 nm for blue can provide particularly good separation with typical Bayer sensors. The green channel is more complicated; around 540–560 nm would theoretically be attractive for the magenta dye, but efficient LEDs in this range are difficult to source, making 525 nm a practical compromise.

Since the release of Jack's scanlight, interest in the subject has grown. Currently (september 2026) the RGB scanlight from film digitization studio Cutenewdesign is commercially available through their website. Tonephotographic is developping their own RGB hardware/software combo. As far as I know, it currently being tested and hopefully will be released in the near future.

This is an important distinction between a general RGB light and a film-scanning RGB light. The goal is not to make the light look like a nice RGB colour to the human eye. The goal is to provide the camera with useful, separable measurements of the film.

The three LEDs do not need to be equally bright. In fact, they generally should not be. The film's orange mask and the different transmission characteristics of the film mean that the camera receives very different amounts of light in each channel. The deep red and deep blue LEDs also interact differently with a Bayer sensor. The practical goal is therefore to adjust the three light levels so that the camera makes good use of its dynamic range in all three channels without clipping. It is recommended to adjust the RGB levels while looking at an unexposed, developed piece of film and aligning the channel histograms. The optimal values are camera-dependent. This is important: there is no universal RGB brightness ratio. The correct ratio depends on the complete scanning system.

My own testing with the Cutenewdesign light paired with my Fuji GFX GFX100S brought following results:

Once we have an RGB scanning light, there are two main ways to use it.

Triple-shot RGB

The first method is to turn on one colour at a time.

Exposure 1: Red light → camera exposure
Exposure 2: Green light → camera exposure
Exposure 3: Blue light → camera exposure

We now have three separate measurements of the same negative. The three images can be combined into a single RGB image. The great advantage is colour separation. The disadvantage is mechanical stability. The camera and film must remain perfectly still during all three exposures. Any movement between exposures can produce colour fringing when the images are combined. This approach can also be automated depending on the RGB light/camera/software combination used.

See my Negpy article for how these three exposures can be combined and processed.

Single-shot RGB

The second approach is to illuminate the film with red, green and blue simultaneously. The camera then makes one normal RAW exposure. This is much faster and eliminates movement between the three colour captures. However, the Bayer sensor still has overlapping spectral responses. So although the illumination itself is narrowband, some crosstalk remains between the camera's RGB channels. The resulting colour channels are therefore not perfectly independent. This can be corrected computationally using a matrix if we characterise the camera and light source. This is the approach used by NegPy's single-shot calibration workflow.

Triple-shot versus single-shot

The two methods can therefore be thought of as two different strategies: Separate the colours physically during capture or take a single RGB mixed separeted (so not white light) correct crosstalk digitally using a custom camera profile calculated from your camera/light source combination.

Neither method is inherently "the correct" one. equential RGB potentially gives the cleanest separation, but requires three exposures. Single-shot RGB is much faster and simpler mechanically, but places greater demands on calibration.

Colour-negative inversion

RGB illumination does not remove the need to invert a colour negative. It changes the quality of the information available to the inversion process. With a conventional white-light scan, the software has to disentangle a mixture of film-dye response and camera spectral response. With narrowband RGB illumination, the capture has already separated the information much more strongly. This can make the inversion process considerably more straightforward and consistend between differing scenes and film rolls. It is one reason why RGB scanning and specialised negative-inversion software such as [NegPy] work particularly well together.

The wider significance

What makes narrowband RGB interesting is that it moves part of the colour problem from software into the physical capture process. Instead of asking software to recover as much information as possible from a broadband measurement, we can design the illumination to produce a more useful measurement in the first place. This is not a completely new concept. It is closely related to how dedicated film scanners and traditional colour-printing systems approach colour separation. What is new is that increasingly sophisticated RGB light sources and open-source software make these techniques accessible to people building their own camera-scanning systems. And because both the hardware and software are still developing, there is considerable room for experimentation. For my own experiments, I have been using a Fujifilm GFX 100S with a Cutenewdesign narrowband RGB light source.