Your Assistance/Advice, Please: Scanning: Resolution v Scaling

[Peter D]

I've recently purchased the HP 4070, mainly because it's new and has a popup
slide/negative adaptor with built-in backlight (I have a large collection of
20-year old slides I want to digitize before they turn to dust). In
attempting to test the scanning quality of various resolutions, I've run
into a snafu of sorts. It may have to do with a lack of knowledge so I
thought I'd run it by the experts here.

It has a max optical resolution of 2400 dpi. The default resolution for
slides is 200 dpi and 300% scaling. . If I scan at the default settings, it
turns out OK. Not great, but OK.

So I decided to scan at 2400 dpi and 100% scaling. The software tells me
that it's pointless, better to scan at 200 dpi with 1200% scaling. So I
saved one at 2400x100% and one at 200x1200%. Both are the same final size
(expected), the same file size (I expected the 2400 dpi one to be somewhat
larger), and both are identical as far as I can see. I even loaded the
images into photo software and blew them up to 400% and they have the same
artifacting and appear to be identical.

Now, here's the problem. It seems counter-intuitive that the lower dpi scan
(at 1200%) is identical to the higher dpi scan (at 100%). After all, If all
we had to do was scale, why even bother with 2400 dpi optical resolution on
a scanner? Also the quality of both final products is a lot less than I had
thought it would be. I'd say "OK" but not what I expected.

Also, should I save as (default) jpg with no compression, TIF, or something
else? I have lots of room for archiving so file size is unimportant.

Any ideas? Thoughts? Advice?
Thanks

 

[Elwood Dowd]

So I decided to scan at 2400 dpi and 100% scaling. The software tells me

that it's pointless, better to scan at 200 dpi with 1200% scaling. So I
saved one at 2400x100% and one at 200x1200%. Both are the same final size
(expected), the same file size (I expected the 2400 dpi one to be somewhat
larger), and both are identical as far as I can see. I even loaded the
images into photo software and blew them up to 400% and they have the same
artifacting and appear to be identical.

They should be identical. 200x1200% = 100x2400 = 300x800%.

I find it easier to think of in terms of pixels. In fact, don't call
them dots or pixels. Call them "bits of information". Various pedants
will tell you that a dot is a speck of ink on paper, and that a pixel is
a picture element in a binary file, yadda yadda. The details are
interesting once you grasp the Big Scheme. For now, concentrate on the
basics.

A 35mm slide is about 1 inch by 1 1/2 inches. At 2400dpi, 100%, the
final result is 2400 x 3200 pixels. If you scan at 300dpi, 800%, you
get.... 2400 x 3200 pixels.

For all of them, the hardware is actually scanning at 2400dpi. It is
simply setting the final file properties differently each way. If you
open each file in photoshop or your pic editor of choice and look at the
image size, you will see that you have either a 1 x 1 1/2 inch picture
at 2400dpi, or a 2 x 3 inch picture at 1200dpi, etc etc. Take the
number of inches and multiply by the number of "dots" (bits of
information) per inch, and that's how big the picture REALLY is.

Higher resolution scanners are expensive because scanning at higher
resolutions is difficult. Scaling in software is NOT the same thing as
actually capturing more information. To prove this to yourself, scan
your slide at 300dpi, 100%, and then again at 300dpi, 800%. Print each
as a 4x6 and see what you get. The 300dpi, 100% version will look all
pixelated because the software had to take 300 x 450 pixels and spread
them across 4 x 6 inches, leaving only 75 bits of information per inch.
For photo-quality reproductions you need at least 240 bits of
information per inch on the paper. Otherwise each bit is 1/75 of an
inch, which I can easily spot even without my glasses. 1/300 is much
tougher without a loupe---even with a loupe. (Maybe I need new glasses.)

The file properties come into play when printing. Inkjet printers
represent a single pixel with an assortment of differently colored dots,
which on some printers are measured in picoliters of ink. That is a
very small amount, much less than the 300 or so bits per inch that you
can see with your eyes. However, the printer needs to print several of
these little guys at a time in one "dot" so your eye will grasp it as a
single color. Printing at higher resolutions gives the printer more
data to chew on but it will actually interpolate up or down as needed to
get the "right" results, as defined by the engineers who built it.

This is one of those things that will plague your brain to no end until
the light comes on. Experimenting helps a LOT. Here is a good place to
go read a lot more about it: http://www.scantips.com

Also the quality of both final products is a lot less than I had
thought it would be. I'd say "OK" but not what I expected.

I am not familiar with the latest HP hardware, but 2400dpi should be
enough to yield fairly decent results, depending on the source material.
Personally I find it very difficult to scan slides and get good
results unless the exposure is spot on. Negative film has much more
latitude with exposure and colors.

If these slides are 20 years old, the colors have likely changed since
they were taken, and it may take some photoshop effort to get them back.
If they were taken with Kodachrome, don't expect them to look
fantastic in any scanner. (My slides are usually badly exposed, faintly
blurry, and my 7-year-old has better composition.)

I suggest reading Wayne's site above, and looking for a good book as
well. If you are a propellerhead, "Real World Scanning and Halftones"
will give you enough to chew on for a year, and backing it up with lots
of experimenting should yield satisfactory results.

Also, should I save as (default) jpg with no compression, TIF, or something
else? I have lots of room for archiving so file size is unimportant.

There is no such thing as JPG with no compression. If you set a very
high quality, you get an excellent reproduction, but JPG by definition
is lossy---that is the nature of its compression algorithm. TIF is
lossless, but can incorporate compression if you want. TIF files can be
huge, but they faithfully record every "bit of information", where JPG
files do not.

If you foresee any intention of editing the photo in the future, it is
very much in your best interest to save as TIF---zip the files or save
with some compression if disk space gets thin, as it will eventually,
and besides smaller files are easier to archive. If the photo looks
great, prints great, colors work and it's sharp enough, and you don't
foresee the need to clone anyone out, etc. go ahead and save as JPG.
The print will look identical to your eyes and the file will take up
less than 10% of the space.

And don't forget to archive onto multiple media! Backups are golden.

Also see earlier thread on "Archival scanning of slides and photos"

Best of luck!

Elwood

 

[Wayne Fulton]

Delwood is very correct, they should be identical.

If you specify 200 dpi at 1200% size, it will scan at 200x12 = 2400 dpi.

If you specify 2400 dpi at 100% size, it will scan at 2400x1 = 2400 dpi.

These are simply two 2400 dpi scans, should be identical.
Images are dimensioned in pixels. Both scans create about 3360x2160 pixels
(assuming roughly 1.4x0.9 inches of film size is scanned)

The only difference is that one image is scaled to print 200 pixels per inch
on paper, and 1200% means to print 12 times larger than the film. And one
image is scaled to print 2400 pixels per inch on paper, prints at original
film size, but printing 2400 dpi on paper is pointless. You want more in the
300 dpi range on paper.

When you specify other than 100% scale, then the resolution you enter is the
printing resolution on paper, and is no longer the scanning resolution.

The scanning resolution = printing resolution x scale factor.

 

[Peter D]

Thank you Elwood (and Wayne) for your explanations. The more I think about
it, the more sense it makes. I have also read through the messages and see
how the "They're _dpi_! No, they're not! They're _pixels_!" debate plays
into this one.

Space is not an issue (I have a Dual Layer DVD burner and a lot of DVDs),
but wanting to "scan once, edit many" is. So I will now set my default file
settings to PNG or TIF/uncompressed -- which one is better, PNG or TIF? Does
PNG lose anything?
I suppose that I can go from there to whatever final output format and file
type I want to.

In regard to the HP software using a default "Slide" setting of 200dpi @
300%, isn't that quite low/shabby? It yields a 4x6 that's only 150 dpi and
an 8x10 that's around 72 dpi. 200dpi x 1200% yields 2400 dpi at original
size, 300dpi at 4x6 and 150 dpi at 8x10. Is that suffuciently high enough
for an 8x10? I'm supposing going higher that any combination that goes above
2400 dpi is pointless

Maybe someone can suggest a "best case" for each source (Color Slide, Color
Negative, Photo). I want to be able to:

a) Print on my own Epson or Canon photo-quality printer and/or burn to CD
and take to photo lab for printing. *Output size will usually be 4x6 or 5x7,
but I might want bigger.

b) Create images for the Web with the option to "click" for a higher-quality
image for viewing more detail and/or printing.

 

[Wayne Fulton]

In regard to the HP software using a default "Slide" setting of 200dpi @
300%, isn't that quite low/shabby? It yields a 4x6 that's only 150 dpi and
an 8x10 that's around 72 dpi. 200dpi x 1200% yields 2400 dpi at original
size, 300dpi at 4x6 and 150 dpi at 8x10. Is that suffuciently high enough
for an 8x10? I'm supposing going higher that any combination that goes above
2400 dpi is pointless

200 dpi 300% scale scans at 600 dpi, which creates about the right ballpark
size image for viewing on the common video screen sizes. HP defaults have no
clue what your purpose might be, but there can be many possible goals, and
yes, printing needs more enlargement.

Full frame 35 mm film width is 24 mm, roughly 0.9 inches is usable. Assuming
that, then 4x6 inches is about (4 inches / 0.9 inches) = 4.4x enlargement,
therefore a 2400 dpi scanned image should print 4.4x at 2400/4.4 = 545 dpi
(which is quite high). 8x10 is about 9x (0.9x9), so 2400 dpi will print it
about 267 dpi, which is right in the ballpark. Again, assuming 0.9 inch film
width, which is not true of other situations.

Maybe someone can suggest a "best case" for each source (Color Slide, Color
Negative, Photo). I want to be able to:

a) Print on my own Epson or Canon photo-quality printer and/or burn to CD
and take to photo lab for printing. *Output size will usually be 4x6 or 5x7,
but I might want bigger.

The source media wont matter to the printing (but for scanning, film is
smaller and so needs more enlargement than prints). The photo labs will
prefer you bring in the ballpark of 300 dpi. They will accept much less, but
you'd like 300 if you can do it. Its a good target for your ink jet on the
good photo paper too.

Meaning, the rule is that to print 4x6 inches at 300 dpi, then you need
(4 inches x 300 dpi) x (6 inches x 300 dpi) = 1200 x 1800 pixels.

Fill in any desired numbers. This need not be exact, but it is the ballpark.

As to video screens (also dimensioned in pixels), few screens are that large.

b) Create images for the Web with the option to "click" for a higher-quality
image for viewing more detail and/or printing.

If some online photo sites do accept a larger image, but show a smaller
resample on the screen, obviously you need to suppply the larger image to
have that choice be available. However regular web sites dont dont offer
that choice. They only have the one smaller size available, too small to
print.

 

[Elwood Dowd]

Output size will usually be 4x6 or 5x7,
but I might want bigger.

For my own archives I aim for final output to be 8x10-ish, 240-300dpi.
That means a file that is near 2000 x 3000 pixels, give or take. That
gives me the option to print up to 8x10 (or smaller, certainly). The
number of photos I print larger than that is extremely small.

b) Create images for the Web with the option to "click" for a higher-quality
image for viewing more detail and/or printing.

Once you have a folder full of "originals" it is very easy to use any
number of tools to automatically create this kind of thing. There is
even a microsoft tool in the XP power tools that does a darn fine job.
I like Irfanview for futzing with batch operations.

Don't bother rescanning for this, it is far easier to batch it, and the
results will be much more uniform.

 

[Peter D]

200 dpi 300% scale scans at 600 dpi, which creates about the right ballpark
size image for viewing on the common video screen sizes.

But I don't care whether the images are being scanned for viewing. The
default is for _printing_ -- that's why I commented that I thought it was
shabby.

If some online photo sites do accept a larger image, but show a smaller
resample on the screen, obviously you need to suppply the larger image to
have that choice be available. However regular web sites dont dont offer
that choice. They only have the one smaller size available, too small to
print.

I'll do it myself, so there is no need to bother with online sites. What
(iyho) is the best resolution/pixel/dpi/whatever it's called for web-based
viewing if a viewer can click and get a better quality image for printing?

 

[Elwood Dowd]

I shoot for 640x480-ish for the large version, or maybe slightly larger.
Anything bigger than that and some people will have to scroll to see
the whole thing.

Small version depends on the page layout, but I rarely have either
dimension bigger than 120 pixels for a thumbnail. That is usually big
enough to see what the photo is about, but small enough to be able to
fit 4 wide on a page.

 

[Don]

Space is not an issue (I have a Dual Layer DVD burner and a lot of DVDs),
but wanting to "scan once, edit many" is.

If that's the case you may consider "scanning raw". That is to say,
use the highest *optical* resolution and color depth. And most of all:
disable any scanning software "extras" such as color adjustments, etc!
In other words, get the best quality the scanner can deliver without
doing any editing at the scanning stage.

Such a scan is also known as a "digital negative". You would then back
that up and work on a copy, reducing resolution and bit depth to
accommodate your desired purpose (printing, web, etc).

If - at a later date - you wish to use the image for other purposes
(e.g. new, higher resolution display or printer) you always have your
digital negative to go back to instead of having to re-scan.

Don.

 

[Elwood Dowd]

If that's the case you may consider "scanning raw". That is to say,

use the highest *optical* resolution and color depth. And most of all:
disable any scanning software "extras" such as color adjustments, etc!
In other words, get the best quality the scanner can deliver without
doing any editing at the scanning stage.

I would recommend setting any major color adjustments in the scanner
software, such as the orange mask when scanning negatives or any other
kind of full mask. Otherwise you are eating up your limited gamut with
information you will later have to throw away. Wayne's site has a great
page on this subject.

The theory of recording raw information is quite sound. In my
experience, though, burning disks with raw images gets old really,
really fast when you realize you are spending time and money recording
data that you will never use. I got tired of it very quickly, anyway!

 

[Skip Allison]

How do you like the 4070? I was looking at that because it will scan 4
slides at once. Does that feature work all right? Salesman said it would
scan 4 slides and save them as 4 files. How well does the individual
picture feeder work for scanning photos?

 

[Don]

I would recommend setting any major color adjustments in the scanner
software, such as the orange mask when scanning negatives or any other
kind of full mask. Otherwise you are eating up your limited gamut with
information you will later have to throw away. Wayne's site has a great
page on this subject.

Yes, removing the orange mask and inverting the negative are standard
operations and there's very little point in skipping it.

The same goes for gamma. Raw scan purists would scan at gamma 1.0 but
since most people edit in a gamma adjusted environment there is no
point in scanning with linear gamma.

The theory of recording raw information is quite sound. In my
experience, though, burning disks with raw images gets old really,
really fast when you realize you are spending time and money recording
data that you will never use. I got tired of it very quickly, anyway!

It all depends on intended use, of course. As I learned Photoshop
better I found I went back frequently. Granted, scanners improve and a
newer one can get more data, but originals don't stay still and
deteriorate with time. So, it's better to get the most while you can.

Also, if you edit images only once and will never go back then - at
first blush - it would appear to make little sense to scan raw.

However, editing images in the tiny preview window and using
underpowered scanner software will never compare to a full blow set of
tools in a dedicate image editing suite.

So even if you don't intend on keeping the originals scanning raw
still makes sense in order to achieve higher quality in postprocessing
(of course, assuming maximum quality is the goal).

Don.

 

[Peter D]

How do you like the 4070? I was looking at that because it will scan 4
slides at once. Does that feature work all right? Salesman said it would
scan 4 slides and save them as 4 files. How well does the individual
picture feeder work for scanning photos?

I answered this one in a new message (in case others were interested or
wanted to contribute)