optimum resolution for repairing old B&W photos?

[Ian Roberts]

Hi

I'm wondering if Ive been overambitious and scanned at a res that is totally
unnecessary.

Ive ran into a few unexpected problems when trying to print some old photos
Ive restored.

Its seems that my system cant cope with the high res (2400dpi) scans Ive
made as I get numerous low memory erros - yet I have 1Gb of RAM!

Ive been working in Corel Photopaint.

The photopaint file is 153Mb. The jpg version of the same image (without
compression) is 108Mb.

When trying to see a print preview on screen, my system seems to grind to a
halt as I have to wait an age for anything to happen. (I have an AMD 3000XP)
Often I have to shut everything down and start again.

Using other imaging software I get the same probs. Even trying to paste a
link into DTP apps I get messages that the file is too big.

The only way I have been able to get around this is to reduce the picure res
from 2400 to 1200dpi.

I'm really surprised. Whats the point of having the option of a scanning res
of 9600dpi or more if the system falls over at 2400!

Do I really need to use 2400??

Thanks for any info.

Ian

 

[Robert Feinman]

Hi

I'm wondering if Ive been overambitious and scanned at a res that is totally
unnecessary.

Ive ran into a few unexpected problems when trying to print some old photos
Ive restored.

Its seems that my system cant cope with the high res (2400dpi) scans Ive
made as I get numerous low memory erros - yet I have 1Gb of RAM!

Ive been working in Corel Photopaint.

The photopaint file is 153Mb. The jpg version of the same image (without
compression) is 108Mb.


Do I really need to use 2400??

Thanks for any info.

Ian

There is no reason to scan at more than 300dpi (some would say 360) and
8 bit depth. The photos don't have more information in them to be
captured.
If they are snapshot sized and you want to go to a 2x enlargement you
can try 600dpi with a small loss of apparent sharpness. This can
sometimes be helped with a little unsharp mask in your photo editor.
Go to http://scantips.com for a good intro.

 

[Journalist-North]

(snip)

Do I really need to use 2400??

Thanks for any info.

Ian

------------

You really need to understand that your printer will throw away anything
more than (generally) ca 300ppi data and go from there. Your print preview
is also trying to fit those large scans into a low res screen with the same
result and by the same method - discarding data. You are telling your comp
to put 10 pounds of s**t into a five pound bag. No wonder it takes ages to
do it.

As for large image files generally... try turning OFF the video
acceleration, but I emphasise that is NOT your problem here... it is your
bloated image files.

Journalist

 

[Bart van der Wolf]

SNIP

There is no reason to scan at more than 300dpi (some would say
360) and 8 bit depth. The photos don't have more information in
them to be captured.

We have no way of knowing if the OP's original prints (may be contact
prints from large format glass plates for all we know) contain 300 or
more ppi of information. Since we can print and see the difference of
a modern inkjet print up to about 720 ppi, I would certainly scan at
at least 600 ppi.

What's more, if the full native resolution of the scanner's sensor is
not used, we effectively skip one or more sensels and scan lines
between the ones we do scan. That will introduce aliasing artifacts
that may or may not become noticeable in the final result (scanning
line art is e.g. more critical than anything else). The best strategy
would therefore be to scan at native resolution or at least at 1200
ppi, and then downsample (done properly by pre-filtering) to 600 or
1200 ppi. If the output must be enlarged versus the original, then the
enlargement factor should be additionally applied when scanning.

That kind of returns us to the OP's problem of memory error
difficulties when handling a 2400 ppi scan. I suggest checking the OS
settings and test the memory banks for faults.

Bart

 

[Gene Palmiter]

I just want to make it clear that you don't scan at 300 ppi 100%. You scan
so that you will have 300ppi at the print size. If you scan a slide at 300
ppi 100% you get about what....500 pixels? If you want an 8x10 then you will
want to scan so that you have 3000 on the long side after you crop.

 

[Atreju]

Ive been working in Corel Photopaint.

I would suggest trying Adobe Photoshop.

I wouldn't touch Corel.

However, this may not be the problem, just a suggestion.


---Atreju---

 

[Kennedy McEwen]

Hi

I'm wondering if Ive been overambitious and scanned at a res that is totally
unnecessary.

Ive ran into a few unexpected problems when trying to print some old photos
Ive restored.

Its seems that my system cant cope with the high res (2400dpi) scans Ive
made as I get numerous low memory erros - yet I have 1Gb of RAM!

What size is the original material that you are scanning? What mode are
you scanning in?

A 2400ppi scan of a 10x8" original print will produce files of the
following sizes depending on the scan format:

8-bit greyscale : 439Mbyte
16-bit greyscale : 879Mbyte
24-bit colour : 1.32Gbyte
48-bit colour : 2.6Gbyte

So it isn't too surprising that you are getting low memory warnings or
errors if the originals are filling the scanner faceplate - although
most modern operating systems should be able to handle such file sizes
without too many problems if it is set-up correctly with the 1Gbyte RAM
you have. Win9x is limited to a maximum of around 1.5GB though, so that
will cause problems.

Ive been working in Corel Photopaint.

The photopaint file is 153Mb. The jpg version of the same image (without
compression) is 108Mb.

Err... unless you are using jpeg2000 then there is no option for
uncompressed files in jpg format, so these file sizes are less than
informative. Save it in tif uncompressed and see the real size.
Irrespective of the jpg file size, the image editor has to expand the
image to the full size to cope with it.

When trying to see a print preview on screen, my system seems to grind to a
halt as I have to wait an age for anything to happen. (I have an AMD 3000XP)
Often I have to shut everything down and start again.

I am not surprised - it would appear you are generating some enormous
files.

The only way I have been able to get around this is to reduce the picure res
from 2400 to 1200dpi.

I'm really surprised. Whats the point of having the option of a scanning res
of 9600dpi or more if the system falls over at 2400!

The system doesn't fall over at 2400ppi scans - it falls over when the
file size gets too big. That is caused by a combination of too high
resolution, too high bit depth or too large a scanned area. If you scan
a small area, say 1" square, then the system will handle 2400ppi fine,
even at 48-bit colour depth, since the file will be around 33Mb in size.
That is what the high resolution of the scanner is for - scanning small
images, such as 35mm film.

Do I really need to use 2400??

Probably not, but it depends on what you are scanning. Photostore
colour photos barely have enough information on them to justify scanning
at 300ppi. Old black and white photos can need 600ppi to get all the
information off the original, but contact prints can have even higher
resolutions, although 1200ppi should be more than adequate. By that
stage further resolution will only reproduce more print emulsion grain,
but no more image content.

Armed with some basic arithmetic, details of the original image size and
a knowledge of your system configuration, you should be able to work out
the maximum resolution and mode that you can safely scan at. With a
little common sense you will most likely find that you *need* a lot less
resolution than the maximum your system can cope with and you can
continue your restoration work without system crashes.

 

[Kennedy McEwen]

There is no reason to scan at more than 300dpi (some would say 360) and
8 bit depth. The photos don't have more information in them to be
captured.

Depends on the original material. A lot of old B7W prints have a lot
more than the equivalent of 300ppi resolution, and contact prints can
readily get up to 1200ppi.

300ppi is a good rule of thumb for commercially printed colour prints,
but that is all.

 

[Kennedy McEwen]

You really need to understand that your printer will throw away
anything more than (generally) ca 300ppi data and go from there.

What sort of junk printer are you talking about that throws away
anything over 300ppi? And even if the operator restricts the print
output to 300ppi, what has that got to do with the scan resolution? I
didn't see anything in Ian's post that indicates he is only interested
in printing at 1:1 scale.

 

[Journalist-North]

What sort of junk printer are you talking about that throws away anything
over 300ppi?

+++ Typically ANY print method [inkjet; laser; offset; webpress; - the
works] that depends on a line screen value (LPI) - whether that value is a
real tangible screen or a virtual one. DPI printer output is NOT the same as
PPI image input - anything the printer doesn't want to use promptly gets
stapled, folded and mutilated in the print driver.

And even if the operator restricts the print

output to 300ppi, what has that got to do with the scan resolution?

+++ The rule of thumb is to scan at the target output size and printer
[device] resolution - that's what. So what is the advantage of making a scan
that produces a 20x40 inch native image size at some God awful resolution,
vs. one that is properly sized for printing at a lower resolution that the
printer can handle? The answer is in the file size. Why on earth scan for a
20x40 and then output to 4x6? As I pointed out you are trying to fit 10
pounds of s**t into a five pound bag - the image is downsampling somewhere
to make it fit, so why not just make the damn thing fit in the first place?

I scan all kinds of films and other media for commercial and in-house
publication/printing daily and the software that I use asks - not what the
original size is, it doesn't care, but what the output print size and
resolution will be. In a corner box it then tells me, OK, I am scanning at
"X" times the original material size and fingures out all by itself, when I
tell it what the output device is, what the "scan" resolution will be to get
the target output size for that device. Sure I can up the resolution
manually to 10, or 15, or 24000 ppi but what's the f**king point if I am
printing full page to an 8.5 x 11 inch piece of paper on a printer that
wants 300ppi at that size output?

didn't see anything in Ian's post that indicates he is only interested in
printing at 1:1 scale.

+++ He's complaining, among other things, about print preview ON SCREEN and
how his comp stalls while trying to fit a massive file into a minuscule "x
by y" screen size box at something less than 100pixels per screen inch. On
the other hand maybe his 100-200 Mb file size would be more appropriate if
he wanted to print a 10 x 20 FOOT billboard sized print - but I doubt it,
because there the line sceen would be in the order of 15LPI (and roughly
100dpi) and only require a scan at the output size with a res of ca 20-30ppi
MAX!

So maybe he doesn't want to print 1:1. If he moves off of a 1:1 print there
IS re-sampling and there is no way around that - up or down. That's exactly
why the target output and output device is the consideration in the initial
scan.

I can not believe that some people treat scanning like their d**k - though I
see it all the time on these forums - bigger is better and the bigger the
better to brag about. Well, I'll tell you what... I had at least one girl
relate to me that she was going out with a guy and in due course things got
romantic but then was confronted with some really massive sized tool on him
and told him flat out "You ain't sticking that thing in me" [end of
romance!] - and that is what the OP's computer is telling him.

 

[Kennedy McEwen]

What sort of junk printer are you talking about that throws away
anything over 300ppi?

+++ Typically ANY print method [inkjet; laser; offset; webpress; - the
works] that depends on a line screen value (LPI) - whether that value
is a real tangible screen or a virtual one.

Rubbish!!!

Many/most laser and inkjet printers can and do produce much more
resolution on the page than 300ppi. Indeed, 300ppi is pretty close to
the worst possible resolution to use (294ppi) with the second most
popular range of inkjet printers on the market! Few, if any, modern
computer printers, are limited to a 300ppi output and they certainly
don't "throw away" anything beyond that. I regularly create certain
types of print at 720ppi (the limiting resolution of my inkjet) and most
of this is actually resolved on the page - as can be seen when viewed
under magnification.

DPI printer output is NOT the same as PPI image input - anything the
printer doesn't want to use promptly gets stapled, folded and mutilated
in the print driver.

You don't say. A little knowledge is indeed a dangerous thing! I used
the term "pixels per inch" quite specifically - because that is exactly
what I meant. Stick around these parts and you might learn enough not
to rely on misconception and rule of thumb!

And even if the operator restricts the print

output to 300ppi, what has that got to do with the scan resolution?

+++ The rule of thumb is to scan at the target output size and printer
[device] resolution - that's what.

Exactly - and did Ian state anywhere in his OP that the output would be
at the *same* scale as the original?

So what is the advantage of making a scan that produces a 20x40 inch
native image size at some God awful resolution, vs. one that is
properly sized for printing at a lower resolution that the printer can
handle?

There are two advantages:

1) it can be printed at a larger scale than the original
2) it overcomes your "rule of thumb" than nothing more than 300ppi can
be achieved on the page.

Whilst the unaided eye may not resolve much beyond the equivalent of
200ppi on the page, there are such things as visual aids which will
increase the perceived resolution significantly.

+++ He's complaining, among other things, about print preview ON
SCREEN and how his comp stalls while trying to fit a massive file into
a minuscule "x by y" screen size box at something less than 100pixels
per screen inch.

No, he is complaining about file size locking up his machine or
consuming available memory. He does not state why he needs such file
sizes and, whilst it may be reasonable to assume that he doesn't, that
would be mere speculation.

 

[JPS]

In message <[email protected]>,

Why on earth scan for a
20x40 and then output to 4x6? As I pointed out you are trying to fit 10
pounds of s**t into a five pound bag - the image is downsampling somewhere
to make it fit, so why not just make the damn thing fit in the first place?

Well, that's kind of extreme, but if the scanner can optically resolve
better than you want to print, it is still a good idea to oversample
*some*, as the scanning process is imperfect but downsampled samples are
closer to perfection.
--

 

[JPS]

In message <[email protected]>,

Many/most laser and inkjet printers can and do produce much more
resolution on the page than 300ppi. Indeed, 300ppi is pretty close to
the worst possible resolution to use (294ppi) with the second most
popular range of inkjet printers on the market! Few, if any, modern
computer printers, are limited to a 300ppi output and they certainly
don't "throw away" anything beyond that. I regularly create certain
types of print at 720ppi (the limiting resolution of my inkjet) and most
of this is actually resolved on the page - as can be seen when viewed
under magnification.

My Canon s9000 resolves to at least 1200 PPI. What I did was create a
complex RGB color, a kind of low-saturation purple that I know couldn't
bypass the color dithering and dither easily, and made a bitmap like
this using the color:

****
****
****
****


****
****
****
****


****
****
****
****


****
****
****
****


I printed this at 1200 PPI, looked at it under a microscope, and the
edges advanced at 1/1200 inch, with the same density at the very edge
for each one.

The idea that inkjets and bubblejets convert everything to 300 PPI or
360 PPI is nonsense. Very few if any actually do that.
--

 

[Kennedy McEwen]

My Canon s9000 resolves to at least 1200 PPI.

That does NOT test the *resolution* that the printer can reproduce - it
merely tests the ink dot placement *precision*. I seriously doubt that
your Canon gets anywhere close to 1200ppi. Whilst I have not personally
tested the Canon s9000, others whom I trust to do so correctly indicate
the limiting resolution is no different from any other Canon inkjet
printer at 600ppi.

To find the actual resolution at which the inkjet resamples the image
prior to dithering, the native and limiting resolution that it can
possibly achieve, find the lowest resolution at which the printer
aliases with full contrast.

To do this, create a test image with 1 pixel wide vertical lines with a
1 pixel wide gap between them. A couple of hundred or so such lines
should be created so use cut and paste or other tools to quickly create
an error free image. Now select half of that line pattern and move it
across the image by one pixel, so that the black lines at the top change
to white half way down the image and vice versa. Copy the section and
paste it into another area of the page, rotating through 90deg.

Now print that at gradually increasing resolution until you can see the
aliasing patterns (thick bands instead of fine lines) appear. Gradually
converge on the resolution at which the alias bands are widest - this
usually produces a black upper and white lower part of the image, but
you may get just a broad grey band, depending on the interpolation your
printer uses. This is *exactly* twice the native resolution of the
printer. When you send this image to the it is downsampling by a factor
of 2 hence, if no interpolation is used, the printer driver selects
either all the black or all of the white lines depending purely on their
positional phase. I suspect that you will find this to be around
1200ppi - indicating that the printer is resampling to 600ppi before
commencing the dither process.

It is impossible for the printer to resolve more than its native
resolution, but it may not even resolve that if the paper is not high
resolution as well.

You can repeat the above test with different coloured lines to confirm
that you have identified the resolution limit independently of the dot
dithering process.

The idea that inkjets and bubblejets convert everything to 300 PPI or
360 PPI is nonsense. Very few if any actually do that.

*All* inkjets, lasers, piezos, bubblejets etc. *DO* resample to som
native resolution before even beginning to dither the ink pattern.