[argyllcms] Re: icclink -G and source gamuts /-profiles

Alastair M. Robinson wrote:

Hi :)

Klaus Karcher wrote:
In Photoshop. I selected ProPhotoRGB as RGB working space, switched to "Custom RGB" (Photoshop adopts the key data from PhroPhotoRGB), changed the Gamma to 1 and saved the setting as ICC profile.
OK. And does your scanner (or camera?) profile have a similar gamma? Or was the idea to use linear RGB for denoising?
No, the input data for my scanner profile has a L* TRC. One ulterior motive to use gamma 1 was indeed to experiment with denoising in linear RGB.
Also, how was your scanner/camera profile generated? I take it it's not an Argyll profile? 

It is an Argyll profile.
Yes, it's the same (I tried AdobeRGB). Even the "Normal Proof" in imgtarget looks completely wrong when using the Gamma 1 source image. The same conversion in Photoshop is ok. 

Interesting.


And you use lcms for the transformations, don't you?
Yes, that's right - so it seems both Argyll and LCMS have issues with a gamma 1.0 input profile. The behaviour of both system improves rapidly as the gamma increases, though - at a cursory glance, LCMS seems fine at 1.1, cctiff without the -p flag seems accurate enough at 1.3 or so (and with -p, of course, is fine even at 1.0)
So what do the input curves of your scanner/camera profile look like compared with RGB profiles with various gammas?
As the scans are pre-linearized to L*, the input curves are nearly linear (but the PCS is Lab, of course). 

Klaus

Other related posts: