PR: Implement support for "Blackmagic Design" camera colour encodings.

[nick-shaw]

Add basic support for Blackmagic camera colour encodings.

[KelSolaar]

Moving some of the discussion here so that they don't disappear in Slack's Void.

So we have three notebooks doing curve fitting:

• @nick-shaw : https://colab.research.google.com/drive/1XNbxO6O6bGjtC1Igzpo7gDE4dtlHPeyW?usp=sharing

This one is the original Nick built with semi-manual break-point and fitting, it assumes a quasi-log function, i.e. colour.models.logarithmic_function_quasilog definition and a linear function.

• @KevinJW : https://colab.research.google.com/drive/1MXUAwuI-9O3tvdvx5NTmGkZtdCGky2kw?usp=sharing

Based on Nick's notebook and implementing fully automatic fitting, it also assumes a quasi-log function, i.e. colour.models.logarithmic_function_quasilog definition and a linear function.

• @KelSolaar : https://colab.research.google.com/drive/1tIdZM9zmcXYEd4IHHJQSgc0lY8PpGGAI?usp=sharing

Based on Nick's notebook and implementing fully automatic fitting, it however assumes the piece-wise CLF camera log function, i.e. colour.models.logarithmic_function_camera.

I would be keen to have a single notebook we can refer to but besides that, we should decide on how we want to implement the functions in Colour. I was hoping that we could use colour.models.logarithmic_function_camera definition, but the last two LUTs, i.e. BMD_POCKET_4K_FILM_LUT and BMD_POCKET_6K_FILM_LUT, have a joining problem which is evident in my notebook when using colour.models.logarithmic_function_camera:

It can be corrected with a blending function but at this point, we might as well use colour.models.logarithmic_function_quasilog definition and a linear segment:

Thoughts?