IDTs: How Are They Designed?

IDT design is often heavily influenced by the specifications of the digital camera system. At the other end of the IDT’s internal pipeline, all other IDT elements being designed, the magnitude of the matrixed, clipped, white-balanced, linearized RGB values for a captured 18% spectrally neutral reflector will decide the value of the final overall RGB exposure scalar. The design of the remaining IDT processes (clipping, minimization of color analysis error and neutral chromaticity difference compensation) all involve engineering decisions and, possibly, proprietary design techniques.

Minimization of Color Analysis Error

The color analysis matrix is derived through linear regression of two sets of colorimetric values: the camera’s linearized, white-balanced RGB values for test stimuli and the expected RGB values of the RICD for the same stimuli. Training spectra selection must balance selection of pure spectral colors with the spectra of real world objects, and of objects measured in isolation vs. objects measured in situ. Simple repetition of colors in the data set is probably the most straightforward way to weight the results, and this technique is sometimes used to boost the priority of neutrals and other important colors, however methods vary based on context and philosophy.

Neutral Chromaticity Difference Compensation

Two engineering approaches are described to go about creating an IDT that compensates for scene-adopted white chromaticity differing from ACES

  1. Chromatically adapting the training colors from the scene adopted white chromaticity to the ACES adopted neutral chromaticity, producing aim colors for the error minimization that have a similar appearance relative to the ACES adopted neutrals the training colors have relative to the scene adopted neutral.
  2. Illuminate the training spectral reflectances by a source that is spectrally similar to CIE Standard Illuminant D60 rather than the actual scene illumination source. This approach produces aim colors for the error minimization that attempt to emulate the capture of the scene under CIE Standard Illuminant D60.

Alternative Design

An alternative procedure utilizes captured images of test charts under an illumination source to create an IDT. The method has been largely used to calibrate camera systems in practice. However, it is highly discouraged in any sort of uncontrolled environments, as it is highly susceptible to contamination of the captured test chart data. It also limits the training
spectra used to those that can be produced using the test chart.