Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.alpha.net!uwm.edu!math.ohio-state.edu!howland.reston.ans.net!news.sprintlink.net!pipex!sunic!news.funet.fi!cs.joensuu.fi!jlavi
From: jlavi@cs.joensuu.fi (Jarkko Lavinen)
Subject: Re: scanner for slides
Message-ID: <1995Feb21.123521.5999@cs.joensuu.fi>
Organization: University of Joensuu
References: <3hurc3$if2@nntp1.u.washington.edu> <3i8jsa$5rt@woodstock.socs.uts.EDU.AU>
Date: Tue, 21 Feb 1995 12:35:21 GMT
Lines: 47

Michael P. Eckert <meckert@ee.uts.edu.au> writes:

>There seems to be some disagreement about whether the Coolscan is a good

>the color correction worries me however.  The grey scale resolution seems
>to be 8 bits per RGB color.  Now the problem is this:  How is optical
>density mapped to the 8 bits per color.  Is there a linear range over

Cannot comment on Nikon, but I can comment my Kodak Eikonix slide
scanner.

It has 4096 element CCD-row and 12-bit ADC. The specs claim S/N ratio
of 1:1000. It also contains onboard LUT to do 12->8 conversion.

Thats about theory. In practise, there seems to be a lot of noise.
Actually so much that one can well can consider it as a 8-bit scanner.

CCD:s have black current noise, which is due to thermal events in each
element. I have tried to analyse the distributions and cannot yet
comment too much. When the scanner is cold, the readouts are around
10 (10ms exposure time) and standard deviation is perhaps 1.5. When it 
warms up, the readout rises to about 30 and distribution gets slightly 
wider, so that maximum deviation from average is about +-6 and even +-10 at
certain area of the row.

Well, one can subtract the avera from the readout. To do this one needs
to know what the black level is and one can measure it. But the black
current rises tenfold with +25 rise in temperature, so one has to track
constantly the behaviour until it stabilises.

If the black current were not enough, the CCD chip itself is noisy.
One might think that with near saturated exposure (readout about 3800 or
so) the readout might be free of dark curren noise. Well then the CCD
chip itself is so noisy that there are even +-40 values from the
average. On the other hand, the distribution of values is concentrated
around the average. It's the odd readouts hear and there which jump
far away from average.

On the other hand, most of the noise comes from CCD and rest of the
electronics is reasonably stable.

I think the scanner manufacturers do not speak too much about these features
unless their model is really good. They cannot afford the customer to
know how much noise and quality compromises there is.

	Jarkko Lavinen
jlavi@cs.joensuu.fi, http://balrog.joensuu.fi/~jlavi/
