Thanks a lot @UltraMo114!

I will take a look in coming days.

@UltraMo114 : As I'm polishing this on a separate branch, I noted that the Hue Composition data is different from that of the paper:

"h_i": np.array([16.5987, 80.2763, 157.779, 219.7174, 376.5987]),

vs

Also, for the computation of $$M$$, $$E_t$$ is not in the numerator:

M_scam = (C_S_ucs * spow(FL, 0.1) / spow(J_scam_safe, 0.27) * et) * Fm_s_param

vs

@KelSolaar Thanks for the careful review and for spotting this.

Please use the values and formulas as they are presented in the paper.

Regarding the formula for M, you're right to point it out. My coding style there wasn't standard. However, it doesn't actually affect the calculation due to the left-to-right order of operations, so the result is correct.

M_scam = (C_S_ucs * spow(FL, 0.1) / spow(J_scam_safe, 0.27) * et) * Fm_s_param

Thanks! What about the chromatic adaptation transform, it seems to be quite different from that recommended in the Comprehensive color solutions: CAM16, CAT16, and CAM16-UCS paper but is not described in the sUCS / sCAM one.

Thanks for the question. You are correct that the sUCS/sCAM paper doesn't detail a new chromatic adaptation transform. That's because it relies on the standard CAT16 transform.

The key reason for the implementation you see is that sCAM is designed to work with tristimulus values (XYZ) under a D65 illuminant.

To handle colors from any illuminant, we use CAT16 as a pre-processing step. The workflow is:

Transform to Cone Space: The input XYZ color is converted to the M16 cone space using the standard CAT16 matrix.

Adapt: Chromatic adaptation is performed in this cone space to the D65 white point.

Transform Back: The adapted values are converted back to the XYZ space. This adapted, D65-equivalent XYZ is then used as the input for sCAM.

This differs from the full CAM16 model, which performs its subsequent calculations directly within the adapted M16 cone space. My implementation is the standard method for applying CAT16 to a model that assumes a specific input illuminant like D65.

Hope this clears things up, and feel free to ask if you have more questions.

It looks very similar to Zhai and Luo (2018) with the addition of the degree adaptation aspect of CIECAM02. I was thinking it should / could be part of its own module in the colour.adaptation package. What should it be named?

You're right, the implementation is conceptually very similar to Zhai and Luo (2018) but incorporates the degree of adaptation from CIECAM02. I agree that it should be its own module within the colour.adaptation package to keep the code organized.

Regarding the name, I would propose one_step_cat16.

My reasoning is that academically, the term "CAT" (Chromatic Adaptation Transform) refers to the general process of adapting from a source illuminant to a target illuminant, without specifying whether the transform is from XYZ to XYZ or from XYZ to a cone space. The work by Zhai and Luo (2018) is essentially an extension of this concept, which could be described as a "two-step CAT."

Therefore, naming our implementation "one-step CAT16" provides a clear distinction from their work and accurately describes its function within the broader context of chromatic adaptation.