Conversion to CIELAB doesn't match other converters about colourful HOT 2 CLOSED

Hi Mike, thanks for reaching out! :-)

I'd guess that the values colormine gives you are wrong.

My process for the conversion you specified is following:

• I assume that the RGB working space is sRGB (note that this can be changed in RGBColor constructor), which is the most commonly used
• I convert it to linear RGB using the inverse companding process defined in the working space - for sRGB I use the process described here
• After I got RGB with linear channels, I convert it to XYZ using a conversion matrix that is constructed from the chromaticity primaries of the sRGB color space, the result will be an XYZ color with reference white point D65 (because sRGB has reference white point D65)
• Then I need to do convert XYZ to LAB color space. But when we talk about LAB color space, we usually mean LAB color space with reference white point D50 (Photoshop uses it as well), so a chromatic adaptation from D65 to D50 needs to be performed
• For the chromatic adaptation, I use the standard Von Kries chromatic adaptation that transforms the colors to LMS color space inside, and I am using Bradford LMS transform matrix (also used in CMCCAT97) for that
• After the chromatic adaptation, I have an XYZ color with reference to D50, which I can convert to LAB using this method

This whole process in Colourful is designed based on the information I collected from Bruce Lindbloom's website with some of my own research of the topic (mainly regarding the existing chromatic adaptation methods).

I tested the input values with this calculator: http://www.brucelindbloom.com/index.html?ColorCalculator.html and the values match. Try entering 0, 1, 0 to RGB, the default settings of that calculator should be the same as default settings of ColourfulConverter class, then the LAB coefficients should match the ones generated by Colourful.

For a second test, try Wolfram Alpha, their values also agree with mine.

My process is very thorough there, and I assume that colormine just multiplies the RGB vector with some hardcoded matrix to get the XYZ values, which might not be 100% correct.

If you face any other issues, feel free to ask :-)

from colourful.

Closing this now. Feel free to reopen for a follow-up.

from colourful.