ðòïåëôù 


  áòèé÷ 


Apache-Talk @lexa.ru 

Inet-Admins @info.east.ru 

Filmscanners @halftone.co.uk 

Security-alerts @yandex-team.ru 

nginx-ru @sysoev.ru 

  óôáôøé 


  ðåòóïîáìøîïå 


  ðòïçòáííù 



ðéûéôå
ðéóøíá












     áòèé÷ :: Filmscanners
Filmscanners mailing list archive (filmscanners@halftone.co.uk)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: filmscanners: Re: Nikonscan v VueScan



I suggest you open Photoshop, open a just-scanned image, bring up the Levels
dialog, change the Option for white point percentage to your chosen value
(hold down the Alt-key and the Auto button changes to say Options).  Set the
black point to 0, if you like.  Click the Auto button.  That's the first
step: Auto Levels.

Then, press the grey dropper, and click on a "desired mid-grey" portion of
the image, and see how the midpoint of the R, G and B channels changes.
That's the second step: gamma.

The only difference I can think of is that the grey dropper sets gamma for
each channel seperately whereas Vuescan's gamma setting sets gamma for
RGB-combined.  Obviously, you can simply type into the Levels dialog an
RGB-combined gamma number instead of pressing the grey dropper and finding a
point on the image that you think should be mid-grey.

Conclusion: the two techniques are strongly related.

"The gamma option in VueScan is equivalent to changing the midpoint of a
correction curve" - not strictly true.  This is because Photoshop's
correction curve, when manipulated at its centre point, does not act like
the gamma function.  Maybe you are referring to the correction curve
function of some other software.  Or maybe you are implying that Vuescan's
gamma is not simply gamma, but gamma and something else.  What am I missing?

Gamma is commutative:  A gamma correction of 1.3 followed by a correction of
0.9 is the same as 0.9 followed by 1.3.  You can multiply the two together
(i.e. 1.17) and make the change in one step (though I can't think of a
reason why you'd do any of this!).

Except that Photoshop's implementation of gamma isn't that accurate, for
some reason...  I think it rounds badly.  Any big change in gamma (e.g. 4
followed by 0.25) will show a noticeable change in the image principally in
the bottom from 0 to 20 on the histogram.  It looks a bit like grain
aliasing!  16-bit data isn't as resilient as you thought!

Here is something quite interesting:

http://www.cs.berkeley.edu/~kobus/research/publications/IST-99-UNCALIBRATED/
IST-99-UNCALIBRATED.pdf - it describes results from neural networks trained
to correct colour casts in images using different training techniques and
contrasts these with more traditional techniques (e.g. white point
determination from maximum R, G, B values).

Ed, there's a challenge for you!

Jawed

> -----Original Message-----
> From: owner-filmscanners@halftone.co.uk
> [mailto:owner-filmscanners@halftone.co.uk]On Behalf Of EdHamrick@aol.com
> Sent: 23 November 2001 08:06
> To: filmscanners@halftone.co.uk
> Subject: Re: filmscanners: Re: Nikonscan v VueScan
>
>
> In a message dated 11/22/2001 8:14:12 PM EST, Jawed@cupidity.force9.co.uk
> writes:
>
> > Setting Color|white point and Gamma is equivalent, in Photoshop
> terms, to
> >  using Auto Levels and the grey dropper.
>
> No, they aren't related at all.  The "Color|White point (%)"
> option controls
> how much of the image clips to white.  This is what most minilabs do
> when printing negative film, and setting this option to 5% or 10%
> duplicates what a minilab does.
>
> The gamma option in VueScan is equivalent to changing the
> midpoint of a correction curve.
>
> Regards,
> Ed Hamrick
>




 




Copyright © Lexa Software, 1996-2009.