Am Freitag, 11. März 2016, 16:25:09 schrieb Denis Testemale:

[...]

> Le 11/03/2016 16:02, Pascal Obry a écrit :
> > 2016-03-11 15:48 GMT+01:00 Denis Testemale <[email protected]>:

[...]

> >> - deactivated all parameter (auto blabla, local contrast, highlight
> >> recovery, etc.) which could modify the luminosity of pixels
> >> - used only the exposure compensation with 0EV, +2EV and +4EV
> >> respectively
> >> - and finally checked visually (which is subjective and prone to errors)
> >> and quantitatively by measuring the RGB values with the sampling tool.
> >> With the 3 softwares, the three images are identical in terms of
> >> exposure/luminosity/RGB channels . I'm not talking about the noise,
> >> obviously. This is logical to me (and to you also I guess :-) ) and
> >> means that the 3 software define the exposure the same way. This was my
> >> point in previous messages when I said that it is an 'absolute' physical
> >> value. I can check, but I'm very confident that it is the same for LR.
> > 
> > But again this is exactly what I did with Lr when I was mapping the
> > .xmp and the exposure
> > was not taken into account the same way. Strange as it is, it is the
> > truth. I would have loved
> > avoiding all this work by just using the Lr values in dt. The same was
> > true with exposure.
> > 
> > I was using Lr on Virtual Box on half screen and dt on the other half
> > for easy comparisons.

Did you also export pictures and sample the pixels? That might be more precise 
than relying on the video stack of a virtualization layer.

> > Note that for exposure the offset is not that large, the mapping is
> > 
> > defined like this:
> >   lr2dt_t lr2dt_exposure_table[] = { { -5, -4.5 }, { 0, 0 }, { 5, 4.5 } };
> > 
> > That is -5 on Lr map to -4.5 on dt. 0 is 0 and 5 is mapped to 4.5 and
> > the mapping is linear.
> > 
> > I don't remember all the details, this is a long time ago but IIRC I
> > was matching the highlights
> > for the 5 vs 4.5, maybe Lr is always doing some hightligh recovery
> > that dt does not.

If that's the reason for the exposure scaling I think we should ignore it. 
Using exposure to compensate for unrelated changes is IMO very bad™.

It seems we need someone with access to Lr to help us debug this for real.

> > Regards,

Tobias

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply via email to