Re: Umax Astra 6x0S - CCD distance - Do they *really* do this that way

becka@rz.uni-duesseldorf.de
Thu, 19 Feb 1998 21:39:56 +0100 (MET)

Hi !

> After this change, all images (except the 4 resolutions above) need to
> be ccd-corrected, but with a float ccd value, -1 < ccd < 1
>
> The only way I find to ccd-correct those images is to add part of the
> current line value for a given channel with part of the above line value
> for the same channel and to use this value.
>
> For example, with a float ccd of .12, I take 88% (100-12) of the current
> line + 12% of the above line for blue channel and 76% (100-12*2) of the
> current line + 24% (12*2) of the above line for the green channel. If
> float ccd is <0, I swap the % values. [This is for Astra 610S, for 600S,
> replace green with red].
>
> >From my point of view, this may lead to reduce image quality, but I
> can't find another way to do this.

There exists an inverse transformation, so no information lost.
Actually I might consider doing this for the MFS12000 alternate correction
scheme, which now only handles full-line distances.

It is maybe a bit less apparent in there, because its lines are only 1/72"
apart (instead of 1/37.5" what your code suggests), but basically what you
are doing is the right thing and it should give the best image-quality you
can reach.

I suppose you know the reason why this happens - just for those who don't:

P\ /R C
A \ / |
P--- x ---G C
E / ^ \ |
R/ | \B D
lens

I.e. The scanner has three CCD lines each with the appropriate filter.
They are offset a little against each other (well - you can't put 2 CCDs
in one place :-) what causes this effect.

Probably any single pass scanner works this way and later corrects the
wrong colors.

CU,ANdy

-- 
= Andreas Beck                    |  Email :  <andreas.beck@ggi-project.org> =
========  GGI - The Right Thing To Do : http://synergy.foo.net/~ggi/  ========

--
Source code, list archive, and docs: http://www.mostang.com/sane/
To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com