How to convert between colour spaces?

[hminle]

Hi Colour Team,

I want to convert from ProPhoto RGB to sRGB
I think of one way to do it: ProPhoto RGB --> XYZ --> sRGB
But the result seems to be wrong

And another thing, I don't know how to choose ILLUMINANT values for the converting function.
Do you guys have any guide lines?

Best,
Hoang

[KelSolaar]

The easiest way to do that is to use the colour.RGB_to_RGB definition, it will compute a transformation matrix with chromatic adaptation if needed (and required by the user).

Note that to get the ProPhoto RGB colourspace class instance, you don't have to do what you did with your first line of code. We do that in the colour.plotting sub-package because the filtering definitions accept regular expressions but in that case you colour access it this way directly:

• colour.RGB_COLOURSPACES['ProPhoto RGB']
• colour.models.PROPHOTO_RGB_COLOURSPACE

Thus to convert from ProPhoto RGB to sRGB you would just do that:

colour.RGB_to_RGB(
    my_array,
    colour.RGB_COLOURSPACES['ProPhoto RGB'],
    colour.RGB_COLOURSPACES['sRGB'])

Cheers,

Thomas

[hminle]

Hi @KelSolaar ,
Thank for your guide
I just wonder: in the RGB_to_RGB function, the output is also based on chromatic_adaptation_transform, right?

[hminle]

Hi @KelSolaar ,
I've just tried your way.
But there're still some points outside the gamut region for sRGB.

Best,
Hoang

[KelSolaar]

It is entirely possible that ProPhoto RGB colours are out of sRGB Gamut, the former is much larger than the latter.

[hminle]

Hi @KelSolaar
So you guys do not apply any Gamut Mapping Algorithm, right?
(https://graphics.stanford.edu/courses/cs178/applets/gamutmapping.html)

Best,
Hoang

[KelSolaar]

Unfortunately no, but I have started to look at that seriously here: https://github.com/colour-science/colour/tree/feature/gamut_mapping, and related issues are #154 and #311.

[sobotka]

Something about that plot is fishy; the chromaticities shouldn’t be clustered along the edge of the locus. Something seems odd there?

[hminle]

Something about that plot is fishy; the chromaticities shouldn’t be clustered along the edge of the locus. Something seems odd there?

@sobotka , as KelSolaar mentioned earlier, they are investigating about the gamut mapping.

[sobotka]

It’s unrelated to gamut mapping.

The chromaticities in the source to destination will remain in relative positions. The issue is the collapsing of the chromaticities along that right edge of the spectral locus.

Is it possible that the ProPhoto image was clipped to ProPhoto?

[hminle]

Hi @sobotka,
I got the image from this link: https://data.csail.mit.edu/graphics/fivek/
The image ID: 0001, expert A.
You can download and check yourself.
The expert edits the raw DNG image using Photoshop and save it with ProPhoto RGB.

I think this process may have some effects on color chromaticities.

Best

[nick-shaw]

I think it may be related to the use of unlinearised data. RGB_to_RGB expects and produces linear data unless apply_cctf_decoding and apply_cctf_encoding are set to True. Likewise I believe the plotting functions expect linear data.