Scanning for greater dynamic range

About this list Date view Thread view Subject view Author view

From: Ken Watson (watsok@frii.com)
Date: 12/02/02-09:48:08 AM Z


I think someone mentioned double pass scanning for higher dynamic range.
This, I believe , is a miss conception.

The issue is how good is the sensor, then the A to D converter then the rest
of the analog circuitry?

To have the sensor operate at it's best, the more light the better until one
of the pixels becomes saturated. I bet no one can find this information out
with respect to say a 18% gray card or anything else.
The problem with most sensors is that they have become relatively small and
can only "store" a relatively small charge for each exposure. I am not
talking about Epson, but there are scanners out there that REQUIRE one
electron to define more than on state of their A to D converter. A silly
issue, but it allows them to advertise" bits of conversion".

If you double scan you still have the same signal to noise ratio, therefore
no increase in dynamic range because you are just summing together to scans
with the same signal to noise ratio. Unless you have software that can
detect noise and remove it one is just fooling self. As far as I know this
does not exist.

Then the question is how good are the specifications of the A to D
converter? Many are specified by throwing away the 2 least significant bits.
This means a 12 bit converter ( 12 X 3 = 36 bits in scanner speak ) is only
really good for 10 bits of accurate conversion. As the bits go up it is
much harder to get accurate conversion to the claimed accuracy. So a 16 bit
A to D converter may have 3 bits of error thus a 48 bit ( 16 X 3 =48 bits ,
scanner speak) really , most likely has 13 bits of accuracy. BUT you end up
storing a lot of noise ( least 3 bits / pixel) plus what ever noise there is
inside the sensor.

Well if you made it this far I expect this is a lot more than most
photographic folks care to know about scanners and their electronics.

Time to stop

----- Original Message -----
From: "Phillip Murphy" <pmurf@bellsouth.net>
To: <alt-photo-process-l@skyway.usask.ca>
Sent: Monday, December 02, 2002 4:25 PM
Subject: Re: Warning: Digital Question

> Hi Sandy,
>
> I use Vuescan and Silverfast and find them both equally useful.
> Genuine Fractals Print Pro 2.5 can be found on the LizardTech site.
> http://www.lizardtech.com
> (MrSid is part of a different package now)
> The need for extending the dynamic range would depend on the negative
material
> used and it's inherent Gamma.
> For an example of situations that would require a greater dynamic range,
try
> Paul Debevec's site: http://www.debevec.org/Research/HDR/
> Use high bit scanning when possible: 16bit mode if you're using B&W
materials.
>
> best regards,
>
> Phillip
>
> Sandy King wrote:
>
> > Phillip wrote:
> >
> > First, thanks for the information, and I have some follow-up
> > questions and comments.
> >
> > >e you'll need to interpolate the data to scale up to 13x19, (I'm
> > >assuming the target is for a 2880dpi or 1440dpi inkjet printer)
> > >use the true optical resolution of the scanner. (2400spi)...
> >
> > That is correct. Will be printing with either an Epson C80 at 2440dpi
> > or with a 2000P at 1440dpi.
> >
> > > If I'm not
> > >mistaken, this scanner is capable of two pass scanning which you
> > >can use to extend the dynamic range.
> >
> > If so I can't find anything about that in the manual. However, since
> > I am scanning negatives, would there be any advantage to extending
> > the dynamic range?
> >
> > > I recommend Vuescan software over
> > >Epson's.
> >
> > I am working with SilverFast, not the Epson Twain software. If
> > Vuescan better than SilverFast?
> >
> > > When you res-up the image in Photoshop or livepicture,
> > >use a final ppi of 360 for the image. (this will give you slightly
better
> > >printer software interpolation).
> >
> > >
> > >Using plug-ins such as MrSid or Genuine Fractals Print Pro (both from
> > >LizardTech, Inc.) to res-up your image will give you a slight advantage
> > >over
> > >Photoshop's engine.
> >
> > Is there a site for downloading these files?
> >
> > >
> > >Needless to say, if your goal is transcribing the negative or
transparency
> > >with accuracy, a film scanner will yield superior results hands down
> > >unless you're
> > >working with something on the order of a Heidelberg flat-bed scanner.
> >
> > I have to say that the quality I am getting from the Epson 2450 with
> > the 5x7 negatives is really very good. Mark Nelson did a scan earlier
> > this year for me from a 5X7 B&W negative using his Imacon Flextight
> > scanner, at about 1600 dpi. I scanned the same negative with the
> > Epson 2450 at the same resolution and frankly if there is any
> > difference in image quality between the 13X19" comparison prints I
> > made from the two scans it is not visible to my eyes. Mind you I am
> > not saying that the 2450 is as good as an Flextight but for this
> > particular application it appears to be.
> >
> > Best,
> >
> > Sandy
> >
> > >
> > >best regards,
> > >
> > >Phillip
> > >
> > >
> > >Sandy King wrote:
> > >
> > >> I have a scanning resolution question.
> > >>
> > >> Situation is this. I am scanning a 4 X 6.5" section of a 5X7" color
> > >> negative with the Epson 2450 flatbed scanner, which has optical
> > >> resolution of 2400 dpi, using SilverFast scanning software as a
> > >> plug-in to PhotoShop. The eventual purpose of the scans is to make
> > >> color prints approximately 13X19" in size, or to make negatives for
> > >> alternative printing. I understand that a maximum resolution of a
> > >> little over 300 dpi at the printing size is needed for either
purpose.
> > >>
> > >> My question is this. What are the pros and cons of of the two
> > >> following scenarios: 1) scanning at 100% of the 4 X 6.5" negative
> > >> size at the maximum optical resolution of 2450 dpi, re-sizing later
> > >> in PhotoShop, or 2) scanning at the desired printing size of 13X19"
> > >> 300dpi?
> > >>
> > >> Sandy King
>
>


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2.1.5 : 01/31/03-09:31:25 AM Z CST