Re: [Gimp-user] Image luminosity

[Elle Stone <ellestone ninedegreesbelow com>]

On 03/10/2018 03:55 AM, Ofnuts wrote:
> On 03/07/18 15:10, WMusc wrote:
> > > Something like:
> > >
> > > Color>Desaturate (luminosity) and check the average or median in the
> > > Histogram dialog.
> > Thanks for the tip. Do you know whether that approach provides the 
> > mean (median,
> > etc.) luminosity value before or after desaturation? It looks to apply 
> > the
> > effect so it's unclear to me. If the program applies it uniformly 
> > across all
> > images I may be able to use it as a proxy for luminosity but would 
> > like to avoid
> > that if possible.
> I expect Color>Desaturate>Luminosity to not change the luminosity of the 
> pixel. The grey value after is normally the luminosity you would have 
> computed from the RGB components you have in the initial image.
>
> This said, in 2018 I'm surprised you are engaging in such manual labor. 
> Find someone (intern?) to write a small program that will extract all 
> these data directly from the image files and create a spreadsheet with 
> the results.

Before finding an intern to write a program, it might be a good idea to 
figure out whether the metrics you've decided to extract are actually 
useful to the task at hand. Garbage in/garbage out, as I've been trying 
to say over in the pixls.us thread on this topic:
https://discuss.pixls.us/t/image-luminosity/6904

As far as extracting information from an image, ImageMagick might be a 
better choice than GIMP - more available metrics, though again it's an 
open question as to whether any of the metrics actually help you decide 
whether an image is "too this" or "not enough that". Remember the red 
square in the middle of an otherwise solid gray image that I posted to 
the pixls.us thread - the relative luminance of the red square is 
exactly the same as the relative luminance of the background, but I'm 
fairly sure just about everyone will say the red square is "brighter", 
as indeed it according to the definition of "brightness" used in current 
color appearance models.