darktable-org / dtdocs

darktable user manual
GNU General Public License v3.0
74 stars 74 forks source link

noise (profiled): confusing description of "adjust autoset parameters" #235

Closed grothesque closed 3 years ago

grothesque commented 3 years ago

The relevant manual section says:

In general, the value of this slider indicates the number of stops by which you had to increase the exposure. For example, if you had to push the exposure 2 stops above what you normally use, then set this slider to 2.

This confuses me to the point that I do not understand what this phrase is trying to say.

The default value of this setting is 1. Taking the example of exposure pushed 2 stops above what one normally uses, I could understand if the slider had to be set to 4 (if its scale is linear), or to 3 (if it's log2).

But 2? That's consistent with square-root-of-linear, but if this parameter indeed works this way, the documentation should be more explicit.

elstoc commented 3 years ago

Thanks @grothesque.

@rawfiner please can you advise on this.

rawfiner commented 3 years ago

Good point. The message conveyed should be 2^(number of stops added), which gives 1 if no exposure compensation is done

grothesque commented 3 years ago

Wow, you guys are responsive...

elstoc commented 3 years ago

Thanks @rawfiner - I'll update it

elstoc commented 3 years ago

@rawfiner - Edit sorry try again

In general, the value of this slider should be set to 2^(exposure compensation) where exposure compensation is the number of stops by which you had to increase the exposure. For example, if you had to push the exposure 2 stops above what you normally use, then set this slider to 4.

?

johnny-bit commented 3 years ago

I wonder if adding another example, eg -1 ev and resulting correction of 0.5 would nail that this is 2^(exposure comp)

matt-maguire commented 3 years ago

Or else we forget about stops and just say that the "effective ISO" used by the denoising algorithm is the actual ISO multiplied by the slider value. The higher the "effective ISO", the more aggressive the denoising parameters will be set.

rawfiner commented 3 years ago

I like @matt-maguire proposal. I think it is clearer