Re: [Gimp-docs] Color Temperature

[Elle Stone <ellestone ninedegreesbelow com>]

On 07/07/2018 02:45 PM, Elle Stone wrote:
> I don't think "What color temperature was the original image" is a 
> well-defined question. If the image was color-balanced by making a 
> neutral object in the image read as neutral, at least in ICC profile 
> software the color temperature of the image is now D50, as the original 
> color cast from whatever light was actually used, has been 
> color-balanced away. But maybe the user actually color-balanced to 
> retain the impression of the color of light in the scene, for example 
> capturing the colors of the sunset instead of white-balancing them away, 
> or to impart a creative look via color-balancing.

Also there's the question of what color balance was set in the camera, 
if the user was shooting jpegs. If the camera was on auto white balance, 
who knows how accurately the image was color-balanced. If on some other 
setting, who know how accurately that setting reflected the actual light 
source.

And if the user was shooting raw, each raw processor has different 
temp/tint numbers for different presets such as Daylight, Incandescent, 
and so forth, and if using "in camera" white balance, the same questions 
as apply to jpegs, also applies to raw files.

Also GIMP doesn't provide "tint" settings, so there's only so much 
GIMP's "Color Temperature" sliders can do. Plus AFAIK GIMP's color 
temperature code is table-based, and personally I don't have a clue how 
table was derived or how accurate it is over what range, or whether/how 
well it applies to color spaces other than sRGB.

Nonetheless GIMP's Color Temperature results are often pleasing and 
believable, even though it's a big guessing game as to what "should" be 
the input temp for any given image.

Best,
Elle


--
https://ninedegreesbelow.com
Color management and free/libre photography