Nikon LS-40 & Vuescan - How many passes are too many?

Nikon LS-40 & Vuescan - How many passes are too many?

Post by Lao Tz » Sat, 15 Mar 2003 01:57:52



Just as the subject reads.  At what point to passes get redundant ?  Is
making 4 passes a waste of time ? or is 2 just as good as 4 ?

How many passes do you make while scanning slides?

Thanks.

 
 
 

Nikon LS-40 & Vuescan - How many passes are too many?

Post by degru » Sat, 15 Mar 2003 10:18:22


16, but most people won't see any difference after 4. i am archiving
some old underexposed slides and in no hurry. Each pass gets you an
extra fraction of a bit assuming the the noise is random. i think i read
somewhere that 16 passes is the equivalent to about 2 bits in noise
reduction.
Frank

> Just as the subject reads.  At what point to passes get redundant ?  Is
> making 4 passes a waste of time ? or is 2 just as good as 4 ?

> How many passes do you make while scanning slides?

> Thanks.


 
 
 

Nikon LS-40 & Vuescan - How many passes are too many?

Post by Kennedy McEwe » Sun, 16 Mar 2003 02:16:37



>> Just as the subject reads.  At what point to passes get redundant ?  Is
>> making 4 passes a waste of time ? or is 2 just as good as 4 ?
>>  How many passes do you make while scanning slides?
>>  Thanks.

Depends on the source material - you can see the benefit of additional
passes with negatives more than with slides.

The benefit of multiscanning is proportional to the square root of the
number of passes.  So doubling the number of passes gives about 1.4x
less noise.  4x gives half the noise or 1 extra useful bit and, as Frank
says, 16x gives a quarter the noise or 2 extra bits.  Just a single
extra bit beyond that needs 64 passes - which is quite a long time and
more passes than Vuescan supports in any case.

Only you can say if the extra scan time is worthwhile since only you can
put a value on your time.

Eventually you might run up against something called 1/f noise which
means you get less than the theoretical halving of the noise for every
4x passes but, assuming that the scanner recalibrates on every pass,
this should not be an issue.  I don't know if Vuescan does this or not
or, if it doesn't, what the 1/f knee is for your scanner to tell you
when this will come into play.  At a guess though, I would expect this
to be somewhat higher than the 16 pass limit so, to all intents and
purposes, it isn't something you need to worry about.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a * when he's pissed.
Python Philosophers         (replace 'nospam' with 'kennedym' when replying)

 
 
 

Nikon LS-40 & Vuescan - How many passes are too many?

Post by Kennedy McEwe » Sun, 16 Mar 2003 02:48:24




Quote:

>Twice is not enough to decide whether a certain pixel really should or
>should not have the value it has. If on the first scan a black area is
>read as 000, and on the second as 255, it would have to be averaged to
>128, which is definitely wrong. Scanning four times, and suggesting that
>the CCD noise is random, the black area would be read as 000, 255, 000
>and 000. Averaging or some smart calculations would indicate that the
>value really should be 000, not 255.

>Four times is the minimum amount of passes to eliminate CCD noise.

With the greatest respect Lauri, that is simply wrong.

Multiscanning exploits Poisson statistics - random numbers - and the
improvement in signal to noise is simply the square root of the number
of passes.  Thus two passes yields a noise reduction of 1.414x, three
passes yields 1.73x and 4 passes yields 2x (or one effective extra bit).

Two is the minimum number of passes necessary to obtain a visible noise
improvement.  Furthermore, that first additional pass makes a bigger
difference to the image noise than any other single additional pass
thereafter, which is obvious from the arithmetic presented above.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a * when he's pissed.
Python Philosophers         (replace 'nospam' with 'kennedym' when replying)

 
 
 

1. Vuescan w/ Nikon LS-40 (coolscan IV) - pixelated images

Vuescan 7.6.28
Windows 2000

Hi all, I have a quick question about my scanning technique.  I have been
scanning slides for a little while now, and I just want to know if there is
a better technique than the one I am using.  Even when I  mess around with
the available filters (IR Clean, Restore Colors, Restore Fading, Grain
Reduction and Sharpen) some of the images seem to be very pixelated.  The
slides are from 1967- 1970, a mix of Kodachrome and regular slides all mixed
up.  Is this something I should expect or correct?

I have read the "Advanced Workflow Suggestions", but I was wondering if
there was anything else I could try.

Thanks,

Matt

Media Type:  Image
Bits per pixel:  64bit RGBI
Scan Resolution: 2900dpi
Auto Focus:  Always
Number of Passes: 4
Long exposure pass: checked
Saving as Raw file.

2. Reading TRSDOS 6 disks on a PC

3. VueScan IR cleaning with Kodachrome slides (Nikon LS-40/IV ED)

4. User Or Refurbished SGI ONXY

5. Vuescan Linux and Nikon LS-40

6. Does MusicKit work under OPENSTEP for Mach?

7. Nikon LS-40 and Vuescan questions

8. removing .bak files

9. VueScan auto crop with Nikon LS-40

10. Success with Nikon LS-40 and vuescan under Linux

11. Vuescan, Nikon LS-40, Fuji NPZ, exposure problems

12. Success with Nikon LS-40 and vuescan under Linux?

13. Minolta Dual Scan Elite II & Nikon IV LS-40 ED