Understand Trichrome capture

Before you start

This page covers three-color digitization of developed film. Each file must come from a separate red, green, or blue exposure; three copies of one image cannot replace them. Ordinary single-shot capture remains an option if you do not have a three-color setup.

How it works

1. Read the film, camera, and reference curves

   Light passes through the film and reaches the camera, which records channel values. The plots show dye density, camera spectral response, and a measurement reference. They describe different things. Overlap alone does not prove color error.

   123

   1. 1
      Dye absorption: effects extend beyond one band

      In the upper box, yellow and magenta both have nonzero values over the same wavelengths. Their main blue/green absorption ranges overlap through side absorption; RGB does not directly measure three pure dyes.

   2. 2
      Camera reception: one band affects several channels

      In the middle box, green and red respond to some of the same wavelengths, so mixed illumination contributes to multiple records. This is normal sensor behavior, not proof of lost information or impossible calibration.

   3. 3
      Measurement reference: calibrate the relationship, not the curve shape

      Status M is a density-measurement reference, not an ideal camera specification. A chart relates input to target; a CCM does not physically reshape sensor curves into this plot. Choose APD or Status M according to the negative workflow.

   Technical notes

   From top to bottom: Dye Density Curves show developed-dye optical density by wavelength; higher density transmits less light at that wavelength. Sensor Sensitivity shows camera-channel reception. Status M Responsivity gives spectral weights for standardized density measurement. λ / nm denotes wavelength. The upper vertical axis is density; the middle and lower axes are responses, not the same quantity. Camera RGB is not a direct reading of three pure dyes, but overlap does not imply that calibration is impossible. Status M is not a universally ideal camera response; choose APD or Status M according to the negative workflow.

   Developed color film records an image through cyan, magenta, and yellow dyes. Cyan mainly absorbs red, magenta green, and yellow blue, with overlapping absorption ranges. Digitization measures the light remaining after passage through these dyes and the base. Changing the illuminant or sensor can change the measurements of the same film. A camera profile for ordinary photography is not automatically a calibration of this film-digitization setup.

   Follow the light: source emission → wavelength-dependent film transmission → weighting by optics, filters, and sensor → a sum in each channel. This additive model applies to black-corrected, unclipped linear signals, not directly to gamma-encoded, graded, or locally processed image values. CFA (Color Filter Array) responses usually overlap. Here spectral crosstalk means mixed illumination contributions, not electrical leakage between pixels. Overlap alone does not prove color error or mean that all three channels carry the same information.

   Optical density describes attenuation: lower transmittance means higher density through a logarithmic relationship. A broadband measurement sums transmitted light weighted by source and receiver, normalizes against a matching no-film reference, then takes the negative logarithm. This generally differs from the correspondingly weighted mean of spectral densities. This is a spectral distinction, not a description of product calculation steps. If the band is sufficiently narrow and transmission varies little within it, the result approaches density at that wavelength—not the density of a pure dye layer. This physical model explains a potential use of narrow bands, not a guarantee that arbitrary LEDs suit every film.

2. RGB lit together: changing the light spectrum

   The three separate peaks illustrate narrow-band illumination: light is concentrated in selected wavelength ranges, changing how dye absorption contributes to the record. The result depends on the light and film; it is not necessarily better than white light.

   

   1. 1
      Narrow bands change the emitted light

      Sampling where one dye absorbs strongly and others less can help distinguish dye changes. Center wavelength, bandwidth, signal strength, and film determine the result; narrower is not always better.

   2. 2
      Dye separation and viewing reproduction differ

      Distinguishing dyes better does not automatically reproduce the film’s appearance under a specified viewing light more faithfully. White light with good calibration remains valid; simultaneous RGB also needs matched calibration.

   Technical notes

   Suitable narrow bands can sample where one dye absorbs strongly and others absorb less, helping distinguish dye changes. Film, wavelength, bandwidth, and noise determine the benefit; narrower is not always better. Dye separation and faithful reproduction of the film under a specified viewing light are different objectives. White light with good calibration can also produce accurate color. With all lights on, a CFA channel may receive several illumination contributions; overlapping curves alone do not establish whether that mixture is difficult to calibrate.

3. Three exposures: one light at a time

   Using one light at a time records the red, green, and blue contributions separately. This refers to black-corrected, unclipped linear signals, not color arithmetic on finished photographs.

   1. 1
      Red only → take R

      Green and blue lights are off. Red may still excite other CFA sites, but their side responses do not supply the composite G and B.

   2. 2
      Green only → take G; blue only → take B

      G comes from the green-only exposure and B from the blue-only exposure. Each record therefore excludes the illumination contributed by the other two lights.

   3. 3
      Separate records do not guarantee more information

      Three shots are not automatically more spectral information. Under ideal conditions, an invertible one-shot mixture yields the same three band values. Practical benefits depend on noise, clipping, bandwidth, and calibration. CFA, dye side absorption, and registration requirements remain.

   Technical notes

   Take R under red only, G under green only, and B under blue only, then compose them. Red can still excite other CFA sites, but those side responses do not supply composite G and B. Each measurement excludes the other two lights, not weighting within the active light’s bandwidth or dye side absorption. Three shots are not automatically more spectral information: with three ideal extremely narrow bands, linear noiseless signals, known responses, and an invertible mixture, one mixed capture can recover the same three band values. Practical benefits depend on noise, clipping, finite bandwidth, calibration, and registration, not the shot count alone.

4. Compare the photographs

   Use the black-cow example to compare two changes: white light versus RGB lit together, then RGB lit together versus three separate exposures. All three use a Bayer camera.

   

   Bayer + white, one captureBayer + simultaneous RGB, one captureBayer + sequential R/G/B, three captures

   1. 1
      Snow: check casts and highlight detail

      Inspect casts and highlight detail without requiring all snow under scene illumination to be colorless white.

   2. 2
      Grass and the cow: check color and shadows

      If separation improves distinguishability and calibration stability, color relationships can improve. These processed photographs are not a controlled accuracy test; greener, darker, or more vivid is not evidence of accuracy.

   Technical notes

   If a setup cannot reliably distinguish dye changes, correction may amplify noise and measurement errors. Suitable bands recorded separately with adequate signal can make those changes easier to distinguish and calibration more stable. This is conditional, not a rule that less spectral overlap always gives better color. Inspect grass/earth distinctions, neutral casts, and hue consistency across tones rather than vividness alone. Three exposures also add time, files, and registration risk. The processed black-cow images illustrate this example, not a guaranteed benefit for every setup.

5. Composition, CCM, and demasking are separate

   The overview arrow only shows three records being combined. Use a matching CCM to correct the capture setup’s color response, then apply negative conversion and viewing. Reversal uses its own positive workflow.

   R · Red captureG · Green captureB · Blue capture

   ↓ Combine in R, G, B order

   1. One linear RGB input frame
   2. Matching input settings and CCM
   3. Negative: roll conversion; reversal: positive processing

   Workflow diagram, not an app screenshot. Composition is neither demasking nor completed color calibration.

   CCM principles: how a chart becomes a correction profile →

   Technical notes

   The arrow in the overview only shows three records being combined. Next, use a matching CCM to correct the capture setup’s color response, then apply negative conversion and viewing. Reversal uses the corresponding positive workflow.

   The records still depend on illumination peaks and bandwidths, channel strength, sensor sensitivity, and dye absorption. A matching chart constrains correction toward the required reference; three shots do not replace calibration. Channel mixtures can be corrected when the input preserves enough distinguishable information. If different spectra produce identical input values but require different target results, a CCM cannot separate them again. Noise, clipping, and mismatched conditions also limit results. Negative density targets and reversal color targets have separate calibrations, neither a universal spectral-recovery tool.

6. Choose Bayer or Monochrome for your sensor

   Choose for the actual camera: Bayer for a color camera, Monochrome for a monochrome camera. The black-cow example uses Bayer. The rgb_7, rgb_8, rgb_9 files are a separate Monochrome example and need their own settings and CCM.

   The two input paths

   Aspect	Bayer	Monochrome
   Sensor	Color-filter-array sensor	Actual monochrome sensor
   Three captures	Each capture is decoded under the RAW input policy before extracting its role channel	Each filtered record supplies its corresponding monochrome channel
   Do not assume	Three captures remove the CFA or decoding interpolation	An unprofiled composite is already standard sRGB/Rec.2020

   Technical notes

   Sensor describes the hardware that produced the files. A monochrome picture style or desaturation does not turn a color sensor into a Monochrome input. A color camera used under three separate lights still belongs to the Bayer path.

   A true monochrome sensor has no Bayer CFA. Each pixel position can measure all three sequential spectral records without Bayer color interpolation. This spatial-sampling benefit is separate from spectral separation through sequential illumination. Bayer cameras can use Trichrome too, but retain their CFA and do not acquire the same per-pixel RGB sampling. Both paths require three acquisitions, more files, and stable registration; movement can create colored edges. Validate one group before committing a roll.