RE: filmscanners: Vuescan long pass mode

["Oostrom, Jerry" <Jerry.Oostrom@Alcatel.nl>]

<break>

> -----Original Message-----
> From: bjs [SMTP:bjs1@home.com]
> Sent: Saturday, February 10, 2001 9:16 PM
> To:   filmscanners@halftone.co.uk
> Subject:      Re: filmscanners: Vuescan long pass mode
> 
> ----- Original Message -----
> From: "Herm" <hermperez@worldnet.att.net>
> To: <filmscanners@halftone.co.uk>
> Sent: Saturday, February 10, 2001 2:20 AM
> Subject: Re: filmscanners: Vuescan long pass mode
> 
> 
> > charge bleeding is a characteristic of CCD sensors, once the electron
> wells get
> > filled up you get these vertical smears as the charge "bleeds" into
> adjacent
> > pixels..its a hardware problem, can only be solved by limiting the
> exposure.
> 
> I agree it is a hardware problem but smart programmers have been working
> around hardware problems for decades.
> 
> I have a Pascal program that takes N files at arbitrary exposure levels
> and
> combines them into one "longpass" result.  It accounts for charge bleeding
> and a number of other issues.  The result has none of the gross errors
> that
> Vuescan currently shows and works far better.
[Oostrom, Jerry]  <phasers locked, fire>
Byron,
"smart programmers" used in comparison to Ed, "gross errors", "far better",
by the nuances in these words you sound as if you feel attacked, just like
Ed and others in reply to you b.t.w. ;-) 
Perhaps, that is the common way we try to make others do what we want, by
shooting them and let them dance to our bullets. In defence they shoot back
with the same attitude. This list has a handful of people who sound like
this in many of their mails and I am astounded that these people are
generally past 40 years of age, i.e. supposedly grown ups. However, they are
also generally the ones quickest to respond to requests for help, so that's
a good reason for me to stay voluntarily on this list.
<phasers down>

<shields down>
Anyway, I'll give it a try, fortunately you all can lough at my silly
attempt to help you and Ed only later, when I have shields up. 

I have no understanding of the idea behind the "long exposure" algorithm,
though I did use it with my Scanwit, resulting in magenta tints in faces,
especially lips, noses and cheeks. It sounds to me as if Ed's algorithm
works like this: you do one or several normal exposure (multi-)scans and one
long exposure scan and combine the results with some weighing factor
(probably fixed or reciprocal to the luminance value of one or all color
channels and also reciprocal to the exposure) for pixels in the long
exposure scan that have not been exposed to the limit (e.g. 255 in 8 bit
scan, only looking at one color channel). 

Perhaps the most simple approach would be to make this weighing factor in
channel X or all channels 0 (zero) for any long pass exposure pixel adjacent
to or in close neighbourhood of a long pass exposure pixel that was
overexposed in channel X. The unfortunate result would be that you loose the
shadow info in areas of high contrast, but the problem of bleeding should be
smaller. In this case you don't make use of knowledge about which pixel is
well exposed and which pixel contains change bleeding, so you don't need to
know.

Another, complex approach would be to apply scouring algorithm in the long
exposure scan to the sections with pixels exposed to the limit (one/all
channels). That is, if scouring algorithm can be applied that way. The
pixels that were exposed to the limit would still have weighing factor 0,
but the adjacent pixels could have their normal weighing factor, whatever
that is.
</shields down>
Time's up.
</break>

I am sure some of you, programmer or not, or Ed, who is a smart programmer
i.m.o., will come up with far more ingeneous approaches and accompany them
even with pseudo code to combat the pixel bleeding problem. Some of you seem
really knowledgeable about the physics of light and the engineerings behind
CCDs and stuff, thus this poses little challenge to at least some of you.
Perhaps inspiration or time available is the limiting factor for most.

Jerry