Scanning objects - any tips?

[Marjolein Katsma]

As long as I don't have a film scanner (and supporting new computer) to
get my photographs into digital form, I'm amusing myself with making
scans with my little HP 3300c flatbed scanner of the objects I brought
from my last trip to serve as illustrations in my travel blog. Well,
mostly it's amusing, but it's also instructive and causing quite a few
grumbles.

Some of these things are easy: things like (hotel) business cards and
admission tickets for various places are just flat and easy enough to
scan. I'm saving those for last. It's the not-so-flat objects that are
more interesting and sometimes really challenging.

For instance there will always be some shadow on and around the object -
that's fine, it may give the effect of a photograph of the object. I was
at first somewhat surprised with the *direction* of the shadow (if the
"top" of the object is near the "top" of the scanner bed, the scahdow is
at the top) but it either doesn't really matter I can just take it into
account. I still haven't quite figured out why the shadow will be in
that direction though.

Yesterday I tackled a couple of bags of snacks (one with a snack still
inside, one empty); one of those gave me a problem: they're made partly
of something like plastic-coated aluminum film: a very shiny surface,
used as the surface to print the text and decorations on. One of them
caused what I think of as "overload" in parts of the unprinted shiny
areas: totally white areas (where one expects some shading) bordered on
one side with very bright green and blue (where shadows should be),
sometimes "radiating" out, which then are bordered by real shadows. One
scan I had to do again, it was so bad.
I found that how hard I press on the scanner cover (and its orientation
- flat or tilted) made some difference but I don't really know how to
control the effect so it's manageable. I was just lucky that the second
scan was a lot better - after wasting a lot of time trying to touch up
the first. I also tried a different tactic with the second scan to touch
up these "overload" areas; once the image size is reduced it's really no
longer visible (so it's "good enough" and a lot faster than my first
attempt but it's still rather a lot of work).

Good practice for using software (PSP 7 now, PSP X purchased but not yet
installed) to touch up an image ;-)

Another factor is that HP's software for this scanner (normally more
than good enough for my flatbed scanning needs) doesn't really offer a
lot of control in this process. I set it at 300dpi and size 100% or 200%
to effectively end up with an "enlargement" of the object which gives me
enough pixels to play with (in PSP) and make cutouts from.

Anyway - do any of you have any experience with scanning 3-dimensional
objects and tips to share? If so I'd be very interested to hear them.

If anyone's interested, I can post my present images online somewhere
(they're not integrated in my blog yet - that needs a new release of my
software which I still have to write).

 

[mp]

I'm glad you posted your experience. Scanning objects is one of the
things that hasn't been fully explored yet.

The "overload" is caused by the light reflecting back into the scanner.
Scanner CCD's don't have a function to control what's sometimes called
blooming in digital cameras.

Try scanning at a low resolution and see if that cuts the blooming.

 

[Marjolein Katsma]

mp ([email protected]) wrote in @o13g2000cwo.googlegroups.com:

I'm glad you posted your experience. Scanning objects is one of the
things that hasn't been fully explored yet.

Hehe - here I come to ask for help and I'm actually contributing? Not
that I mind, but it's funny.

The "overload" is caused by the light reflecting back into the
scanner.
Scanner CCD's don't have a function to control what's sometimes
called blooming in digital cameras.

Well, I could figure out what was happening - just not how to control
it. Actually I wonder whether it might damage the CCD? I don't really
have experience with digital cameras and have never seen it before. Are
you suggesting they do have a way to control the effect?

What is curious (to me) is that the effect is directional, too (as with
the shadows) but this is horizontal. I'm guessing it has something to do
with the sanner's mechanics but I don't know how.

Try scanning at a low resolution and see if that cuts the blooming.

Good idea. I immediately tried that, the last sample was still on the
scanner bed. ;-) I set the resolution back to the default (150dpi) and
tried to keep the scanner lid in the same almost horizontal position I
had it with the last scan.

Given that the effect is variable and the position of the scanner lid
may (or may not) influence the amount of overload (light scattering?)
it's hard to judge. But if anything the new scan at lower resolution is
worse rather than better (than my last scan at 300dpi).

I've uploaded two small samples here (unedited cutouts, saved as jpg
with "no" compression) :

http://www.xs4all.nl/~iamback/tests/scanning/

overload1.jpg and overload2.jpg both show how I'm getting nice realistic
rendering of the shiny material where it's printed, but the unprinted
silvery border has the overload. In the first sample you can see how
that "bleeds" into the background (in an earlier scan that was like a
long, sharp spike). To the eye the green and blue look solid, but that's
because they are so highly saturated; actually there is a whole range of
slightly different hues for each (the green and the blue). Realizing
this, I hit on trying a correction by simply reducing the saturation in
those areas, because the shading on the silvery stuff in the normal
areas is gray anyway. That worked far better than my first approach of
"color replacing" individual colors.


So, given that there is no real improvement (if any) with the lower
resolution, I think I'll stick with the higher resolution which just
gives me more pixel material to work with to retouch the image in PSP
(_before_ reducing it to a publishable size).

Any other suggestions welcomne, of course!

 

[Marjolein Katsma]

Marjolein Katsma ([email protected]) wrote in

I've uploaded two small samples here (unedited cutouts, saved as jpg
with "no" compression) :

http://www.xs4all.nl/~iamback/tests/scanning/

overload1.jpg and overload2.jpg both show how I'm getting nice
realistic rendering of the shiny material where it's printed, but the
unprinted silvery border has the overload.

Suddenly hit on the idea that the white scanner lid might have some
effect, so I also tried with a black (fleece) sweater thrown over the
sample. If the scanner software would "correct" for the total amount of
light (I don't know, really) I would have expected worse overload in the
shiny areas. Instead, it seems somewhat better (but I don't really like
the black background as it also gets rid of the subtle shadow I had).

Sample overload3.jpg

 

[CSM1]

Marjolein Katsma ([email protected]) wrote in


Suddenly hit on the idea that the white scanner lid might have some
effect, so I also tried with a black (fleece) sweater thrown over the
sample. If the scanner software would "correct" for the total amount of
light (I don't know, really) I would have expected worse overload in the
shiny areas. Instead, it seems somewhat better (but I don't really like
the black background as it also gets rid of the subtle shadow I had).

Sample overload3.jpg


--
Marjolein Katsma
*Help with HomeSite/Studio: http://hshelp.com/
*Travel blog: http://blog.iamback.com/
*Spam reporting addresses: http://banspam.javawoman.com/report3.html

You have already tried a black cover, and seen that helps. But you need more
adjustment.

Maybe you can adjust the white and black points in the histogram. I am not
familiar with the HP software so I can not tell the steps involved.

All scanners have some sort of "Pro" mode where you can adjust the scanning
parameters.

To get an idea what I am talking about, do a little reading on getting
better scans.
http://www.scantips.com/simple.html

The home page and start is:
http://www.scantips.com/

 

[humanoid]

As long as I don't have a film scanner (and supporting new computer) to
get my photographs into digital form, I'm amusing myself with making
scans with my little HP 3300c flatbed scanner of the objects I brought
from my last trip to serve as illustrations in my travel blog. Well,
mostly it's amusing, but it's also instructive and causing quite a few
grumbles.

Some of these things are easy: things like (hotel) business cards and
admission tickets for various places are just flat and easy enough to
scan. I'm saving those for last. It's the not-so-flat objects that are
more interesting and sometimes really challenging.

I know a woman, a jewelry maker, who scans all of her work. She
drapes a dark opaque cloth over the jewelry instead of using the scanner
lid. It works very well. Make sure it is a finely made cloth, like a silk
scarf, so it will drape over the object and not just cover it.

 

[Don]

Well, I could figure out what was happening - just not how to control
it. Actually I wonder whether it might damage the CCD? I don't really
have experience with digital cameras and have never seen it before. Are
you suggesting they do have a way to control the effect?

The effect is called "blooming" (see below).

It will not damage the CCD and the quick way to control it, in theory,
is to scan at an angle so the reflections don't hit the CCD array head
on. However, since flatbeds have a very narrow depth of field that's
easily said than done. On top of focus issues, having the object at an
angle will also cause spatial distortions.

What is curious (to me) is that the effect is directional, too (as with
the shadows) but this is horizontal. I'm guessing it has something to do
with the sanner's mechanics but I don't know how.

Flatbeds work with reflected light. The catch is you can't have the
light source (the lamp) and the CCD at the same angle relative to the
object i.e. in the same place (the ideal situation). So the light is
at an angle and this causes the shadows.

Now that's bad enough, but it gets worse. The RGB filters are also at
an angle relative to each other. This causes "color shadows" but you
really have to look hard to see them

Unlike Vuescan users ;o) I do look (very!) hard and examine the images
in detail, so:

I noticed this when scanning (thick B&W) photographs. There's a very
narrow blue cast at the leading few pixels and a red cast at the
trailing end. Its intensity depends on the photo but it's always
present. Although, sometimes, you may need to look very hard.

Given that the effect is variable and the position of the scanner lid
may (or may not) influence the amount of overload (light scattering?)
it's hard to judge. But if anything the new scan at lower resolution is
worse rather than better (than my last scan at 300dpi).

Of course it is! ;o) Anything other than native resolution will be
worse.

When you scan at lower than native resolution you let the scanner
driver reduce this resolution. This may be done in a number of
different ways but whichever one it is, it's bad. Why? Because the
scanner driver favors speed over quality.

Therefore, if quality is the goal, it's always better to scan at
native scanner resolution and reduce afterwards in an image editor.

Any other suggestions welcomne, of course!

What you are really seeing is called "blooming" i.e. an area of the
picture is overexposed. This causes the neighboring pixels (CCD wells)
to be overwhelmed and you will also notice "bleeding" on the CCD
(horizontal) axis if you look closely.

Blooming can be cured by twin scanning and combining of the two
images. Of course, the problem is flatbeds don't usually have exposure
control.

But scanning reflective objects is always going to be a challenge just
like taking pictures of reflective objects is a challenge e.g. sun
reflections against water ripples. It's the same thing, really. It's
like taking a picture of yourself in a mirror with a flash! ;o)

Don.

 

[catfish]

and try turning out room lights, drow the blinds, and reduce other
external light sources as much as you can.

 

[Marjolein Katsma]

Don ([email protected]) wrote in

The effect is called "blooming" (see below).

It will not damage the CCD

Ah, that's a relief.

the quick way to control it, in theory,
is to scan at an angle so the reflections don't hit the CCD array head
on. However, since flatbeds have a very narrow depth of field that's
easily said than done. On top of focus issues, having the object at an
angle will also cause spatial distortions.

Well, I'm talking about (somewhat) three-dimensional objects here, so
there are always bits off the glass. They are just slightly out of
focus, not too much to completely blur the decoration and text on teh
package (the text stays readable), but enough to actually make it look
3D. So that's fine.

But "angle" can be several things. For now I've just aligned the
rectangular bag with the rectangular glass plate. May if I turn it a
little I could get less blooming. I'll try that next!

Flatbeds work with reflected light. The catch is you can't have the
light source (the lamp) and the CCD at the same angle relative to the
object i.e. in the same place (the ideal situation). So the light is
at an angle and this causes the shadows.

The shadows are fine, actually (just at the other end than I had
expected). The problem with the black (fleece) cover is actually that I
_don't_ have the shadows any more; the blooming is only slightly less.
So that's a trade-off but I think for these bags at least I prefer the
shadow and doing a little more work to retouch the blooming.

Now that's bad enough, but it gets worse. The RGB filters are also at
an angle relative to each other. This causes "color shadows" but you
really have to look hard to see them

Unlike Vuescan users ;o) I do look (very!) hard and examine the images
in detail, so:

I noticed this when scanning (thick B&W) photographs. There's a very
narrow blue cast at the leading few pixels and a red cast at the
trailing end.

Now this is really interesting. I scanned part of a black cap (with logo
and text printed on it - that's the part I really wanted and). Of course
the black is not totally black but you get some light reflecting off the
threads in the material. I noticed a definite red cast in the lighter
areas and also (much less) a blue cast. I played with curves (in PSP) to
reduce that so it looks a more natural black - but it took quite abit
of fiddling. That was a few days ago.

This morning, I scanned a bit of a jacket made of a shiny black fabric
with little red flowers woven in. Enlarging my scan I noticed the very
same red and blue cast on the black material!

So, I know exactly what you're talking about - but correcting it is
another matter. I hope I saved that adjustment layer I used on the cap -
if so I could try copying it and applying it to the jacket detail as
well; it could be at least a starting point.

Of course it is! ;o) Anything other than native resolution will be
worse.

Well, the scanner sets the deafult at 150dpi (good enough for most
purposes I use the scanner for), but it allows you to change the
setting; I often bump it up to 300; on the page where you do that,
there's also a checkbox to allow a "rescan" for the higher resolution
(reducing the first pass to a "preview").

If you leave it off, and save the scan, it will just save what it
already has. It you check it though, it will rescan before saving and
- going by the sound and movement of the light - it *seems* to do a sort
of dual scan by going over sections twice. I don't know how to describe
it better. My guess is that it uses the dual scan to get the higher
resolution in this case, and if you don't allow a rescan it just does an
interpolation of what it already has (which I guess is at 150dpi).

But you can *also* set the image size. Here I get confused about what
the scanner software does (no option to rescan for that setting) but I
suspect it's only interpolation. But this does give me more pixels to
work with and sometimes it helps.

Therefore, if quality is the goal, it's always better to scan at
native scanner resolution and reduce afterwards in an image editor.

So what I do (often) is 1) set resolution at 300dpi (instead of 150)
and with the checkbox on this does change the scanning behavior at the
hardware level; and 2) (optionally) bump up the size to 150% or 200%.
That gives me more pixels to edit before (ultimately) reducing the size
again in the editing program (PSP).

Any flaw in that approach?

What you are really seeing is called "blooming" i.e. an area of the
picture is overexposed. This causes the neighboring pixels (CCD wells)
to be overwhelmed and you will also notice "bleeding" on the CCD
(horizontal) axis if you look closely.

Right - it's "horizontal" - sometimes even with considerable "spikes"
(all in one direction). No need to look closely: it's quite obvious.

Blooming can be cured by twin scanning and combining of the two
images. Of course, the problem is flatbeds don't usually have exposure
control.

None of that (nor histograms or contrast or whatever) on mine.

But scanning reflective objects is always going to be a challenge just
like taking pictures of reflective objects is a challenge e.g. sun
reflections against water ripples. It's the same thing, really.

"Challenge" is right. But the more I try scanning these things I
brought, the more fun I'm having. Hardly anyone will suspect the amount
of work that went into the illustrations once they're on my site but
that's OK; and meanwhile I'm learning a lot. (No problem living up to my
motto of "learn something new every day" these days.)

 

[Marjolein Katsma]

humanoid ([email protected]) wrote in

I know a woman, a jewelry maker, who scans all of her work. She
drapes a dark opaque cloth over the jewelry instead of using the
scanner lid. It works very well. Make sure it is a finely made cloth,
like a silk scarf, so it will drape over the object and not just cover
it.

Oh, nice! So I hit on the right idea by throwing over my black fleece
sweater! The only problem with that is that it doesn't give me the shadows
which (at least for some objects) can actually be nice.

 

[Marjolein Katsma]

CSM1 ([email protected]) wrote in

You have already tried a black cover, and seen that helps. But you
need more adjustment.

Maybe you can adjust the white and black points in the histogram. I am
not familiar with the HP software so I can not tell the steps
involved.

All scanners have some sort of "Pro" mode where you can adjust the
scanning parameters.

Not all! Mine doesn't. That's why I mentioned the scanner software gives
me very little control: there _are_ some controls but nothing that
"advanced". Before I embarked on this little project I downloaded and
installed the latest version of the software (already two years old -
but the scanner is even older): HP PrecisionScan LT. Histograms? No such
thing. (One reason I was hoping I could use VueScan with this thingy -
but I can't.)

http://www.scantips.com/

Yes, great site. I've poked around there quite often already.

Actually, http://www.scantips.com/simple1.html (the next page after
http://www.scantips.com/simple.html) states:
"The HP scanners generally do not provide this standard tool (however
PrecisionScan Pro has it). It is not that some scanners do not do this,
but instead that they do this contrast adjustment automatically in
software, and manual control is not available to the user."

I could try adjusting the histogram in PSP but then of course the
blooming has already occurred. For these particular scans of the bags of
snacks my method of saturation adjustment on just the areas with the
blooming works just fine - and the rest of the scan really is quite nice
(given what I'm working with) with nice texture of the shiny material
and enough sharpness to read the text quite clearly. Obviously what's
away from the glass plate will be out of focus, but that's OK, it's only
slightly out of focus; that actually adds to the 3d effect.

 

[Maris V. Lidaka Sr.]

Unfortunately, the other option - Silverfast - doesn't support your scanner,
either.

Try, perhaps, a while cloth cover, which would be more snug than the hard
scanner cover.

Maris

 

[Marjolein Katsma]

Maris V. Lidaka Sr. ([email protected]) wrote in [email protected]:

Try, perhaps, a while cloth cover, which would be more snug than the
hard scanner cover.

Yes, good point. I'm already experimenting with various cloth covers
(putting my collection of scarves to good use). And in deed (Don!) I found
that turning the object slightly on the glass can help to combat the
blooming, mostly by "directing" it where it will do least damage to teh
image and is easier to retouch.

It's a time sink, but I'm getting experience and it's really fun now.

 

[Don]

But "angle" can be several things. For now I've just aligned the
rectangular bag with the rectangular glass plate. May if I turn it a
little I could get less blooming. I'll try that next!

No, what I mean is angle relative to the glass itself i.e. have one
part of the bag touch the glass while the opposite end is lifted a
bit. The idea is to have the reflections point away from the CCDs.

Of course, that's not easy because the bag is bound to be crumpled and
there will always be some parts which will reflect the light into the
CCDs. (However, that approach could work with flat objects.)

On the other hand, the crumpled state is why rotating the bag on the
glass can also result in some reflections being eliminated.

The shadows are fine, actually (just at the other end than I had
expected). The problem with the black (fleece) cover is actually that I
_don't_ have the shadows any more; the blooming is only slightly less.
So that's a trade-off but I think for these bags at least I prefer the
shadow and doing a little more work to retouch the blooming.

Actually, you do have the shadows, but you don't see them because of
the non-reflective back.

Actually #2, strictly speaking, there are two types of shadows. The
shadows resulting from different angles between the light source and
the CCD (that will remain) and shadows due to light reflected from the
cover (that can be eliminated with a non-reflective back).

BTW, instead of a non-reflective back you can simply leave the scanner
lid up and you'll get the same result. The light around the object
will just go straight up and not be reflected. If you do that, reduce
ambient light to avoid it hitting the CCDs inadvertently.

BTW #2, lifting the lid and doing a scan of "nothing" is actually a
very good way to check if your glass is dirty (or bad!). The scan will
appear pure black but if you brighten it up in an image editor you'll
see every dust speck and fingerprint. It's quite an eye opener to do
that although it causes some people to scream with horror! ;o)

.... cut ...
So, I know exactly what you're talking about - but correcting it is
another matter. I hope I saved that adjustment layer I used on the cap -
if so I could try copying it and applying it to the jacket detail as
well; it could be at least a starting point.

The way I correct this is to go straight to individual channels. You
can do a number of things from manual to various gradients, etc.

Well, the scanner sets the deafult at 150dpi (good enough for most
purposes I use the scanner for), but it allows you to change the
setting; I often bump it up to 300; on the page where you do that,
there's also a checkbox to allow a "rescan" for the higher resolution
(reducing the first pass to a "preview").

If you leave it off, and save the scan, it will just save what it
already has. It you check it though, it will rescan before saving and
- going by the sound and movement of the light - it *seems* to do a sort
of dual scan by going over sections twice. I don't know how to describe
it better. My guess is that it uses the dual scan to get the higher
resolution in this case, and if you don't allow a rescan it just does an
interpolation of what it already has (which I guess is at 150dpi).

I'm not familiar with that scanner/software but that sounds exactly
right.

But you can *also* set the image size. Here I get confused about what
the scanner software does (no option to rescan for that setting) but I
suspect it's only interpolation. But this does give me more pixels to
work with and sometimes it helps.

Could be the dpi setting which is only relevant for printing?

So what I do (often) is 1) set resolution at 300dpi (instead of 150)
and with the checkbox on this does change the scanning behavior at the
hardware level; and 2) (optionally) bump up the size to 150% or 200%.
That gives me more pixels to edit before (ultimately) reducing the size
again in the editing program (PSP).

Any flaw in that approach?

Well, there are a number of things to keep in mind. First of all, what
is the native (hardware) resolution of your scanner? What I mean by
this is how many elements are there in the CCD array (the horizontal
resolution) and what is the rating of the stepper motor (the vertical
resolution). That is your native resolution.

However, scanner manufacturers include interpolation and pretend to
have resolutions up to 9600 and more. All that is fantasy. If you take
a native resolution scan and increase resolution in an external image
editor you'll get far better results because of superior algorithms.

The reason it's important to know the native resolution is to be able
to choose optimal scanning resolution. You should only use integer
divisions of native resolution e.g in case of a 1200 scanner use 600
(1200/2) or 400 (1200/3) or 300 (1200/4) etc. Any other resolution,
e.g. 350, will case some sort of interpolation.

I don't know how your scanner software handles the 150% or 200% size
increase but my instinct is to be very suspicious that's a fudge of
some sort (either interpolation or merely a dpi change). If you get
more pixels it appears to be interpolation, though, but I don't know.

None of that (nor histograms or contrast or whatever) on mine.

All that is image editing and since there's no exposure control on a
regular flatbed, a histogram doesn't really help either.

"Challenge" is right. But the more I try scanning these things I
brought, the more fun I'm having. Hardly anyone will suspect the amount
of work that went into the illustrations once they're on my site but
that's OK; and meanwhile I'm learning a lot. (No problem living up to my
motto of "learn something new every day" these days.)

Absolutely! And it's a lot of fun to experiment and try unusual
things. But you're right about some people not appreciating the amount
of work and effort that goes into certain things. I myself value such
attention to detail.

On the other hand, there are also people who are easily fooled and
admire inferior results. But I went this far without mentioning
Vuescan, so I better stop now... ;o)

(A bit of self-deprecating humor, there... ;o))

Don.

 

[Maris V. Lidaka Sr.]

A good book for your projects - scanning objects rather than just images -
might by _Start with a Scan_ by Janet Ashford and John Odam

http://www.amazon.com/gp/product/0201710978/104-8109926-1917511?v=glance&n=283155&s=books&v=glance

or TinyURL http://tinyurl.com/dnae9

Maris

 

[Marjolein Katsma]

Don ([email protected]) wrote in

No, what I mean is angle relative to the glass itself i.e. have one
part of the bag touch the glass while the opposite end is lifted a
bit. The idea is to have the reflections point away from the CCDs.

Yes, I got that - but these 3D objects have their own shape - and
consequent angles.

Of course, that's not easy because the bag is bound to be crumpled and
there will always be some parts which will reflect the light into the
CCDs. (However, that approach could work with flat objects.)
Precisely.

On the other hand, the crumpled state is why rotating the bag on the
glass can also result in some reflections being eliminated.

Right, your story about angles just triggered the thought that rotating
the object on the glass would at least change the angle with respect to
the light tube. It works a little indeed, I can more or less "direct"
the blooming into areas where it's easier to touch up.

No shiny completely flat objects so far, but I'll keep it in mind.

BTW, instead of a non-reflective back you can simply leave the scanner
lid up and you'll get the same result. The light around the object
will just go straight up and not be reflected. If you do that, reduce
ambient light to avoid it hitting the CCDs inadvertently.

I'll try that next. I've been experimenting with various cloths (scarves
and pieces of clothing!).

BTW #2, lifting the lid and doing a scan of "nothing" is actually a
very good way to check if your glass is dirty (or bad!). The scan will
appear pure black but if you brighten it up in an image editor you'll
see every dust speck and fingerprint.

Ah, yes, I had already noticed there seemed to be a few specks. Now
where was that thread about cleaning the glass plate? I'll want to clean
the inside of the cover, too, since when I do use the shadows the
background should be clean as well. (I'm pretty good with the clone
brush but anyway. ;-))

The way I correct this is to go straight to individual channels. You
can do a number of things from manual to various gradients, etc.

I did that for the black fabrics to get rid of the red and blue casts.

But correcting in individual channels is where I'm struggling now - the
scanner seems to always have a slight reddish cast. I'm working on a
(Chinese) bag of what is misleadingly called "American fried beans"
(actually something like "wasabi peas" if that means anything to you).
The peas are bright green but somehow they come out brownish - and I'm
damned if I can correct it to the appropriate color without having a
negative influence on all the other colors. Trying histograms as well
as curves for individual channels (and combinations) but I just cannot
get it completely right (albeit a lot better than the actual scan).

I'm not familiar with that scanner/software but that sounds exactly
right.


Could be the dpi setting which is only relevant for printing?

Well, the dpi setting is what *allows* the "dual scanning", size setting
does not. So I don't think dpi is for printing only - if you check the
setting for rescanning for higher resolution it actually results in more
pixels and more detail.

I don't print anyway, so pixels are what counts (and "inches" are
utterly meaningless). But then size also causes more pixels - but since
there's no rescan involved my guess is that's purely interpolation
(which can help but doesn't necessarily).

First of all, what is the native (hardware) resolution of your
scanner? What I mean by this is how many elements are there in the CCD
array (the horizontal resolution) and what is the rating of the
stepper motor (the vertical resolution). That is your native
resolution.

I wish I knew. I googled a bit and came up with this:
http://www.anyweb.co.uk/computers/listings/242.html
If correct, it would be 600dpi (then I'd deduce only the preview is
actually 150 but you have to tell it to rescan to avoid interpolation
for 300 or 600 dpi). But I'm convinced there is some dual scanning
taking place when not using the default 150dpi and I don't know where
that comes into the picture with these resolutions (the original driver
I used was in Dutch and it had the curious "dubbel beeld" in its title
("double image") - which matches my observation that it rescans in
sections).

However, scanner manufacturers include interpolation and pretend to
have resolutions up to 9600 and more. All that is fantasy. If you take
a native resolution scan and increase resolution in an external image
editor you'll get far better results because of superior algorithms.

I'd certainly not believe 9600 but I'd believe 600 (with rescanning). I
could certainly try enlarging in PSP instead of changing the size in the
scanner driver and see if that makes any difference.

You should only use integer
divisions of native resolution e.g in case of a 1200 scanner use 600
(1200/2) or 400 (1200/3) or 300 (1200/4) etc. Any other resolution,
e.g. 350, will case some sort of interpolation.

That makes sense, thanks.

I don't know how your scanner software handles the 150% or 200% size
increase but my instinct is to be very suspicious that's a fudge of
some sort (either interpolation or merely a dpi change).

Like I said, I suspect it's merely interpolation.

On the other hand, there are also people who are easily fooled and
admire inferior results. But I went this far without mentioning
Vuescan, so I better stop now... ;o)

Ah, finally - you had me worried for a moment... :^)

 

[Marjolein Katsma]

Maris V. Lidaka Sr. ([email protected]) wrote in

A good book for your projects - scanning objects rather than just
images - might by _Start with a Scan_ by Janet Ashford and John Odam

Thanks for the hint. I might save up for that with Amazon's Mechanical
Turk! ;-)

 

[Marjolein Katsma]

Don ([email protected]) wrote in

BTW, instead of a non-reflective back you can simply leave the scanner
lid up and you'll get the same result. The light around the object
will just go straight up and not be reflected. If you do that, reduce
ambient light to avoid it hitting the CCDs inadvertently.

Oh, what fun! I tried that and ran smack into a new problem (or rather,
recurring problem that I tought I had a good workaround for):

The HP scanner software for this machine *automatically* attempts to
straighten a scan - using a really inferior technique instead of (for
instance) PSP's much superior resampling when rotating an image.

The workaround I found was to
1) set "view -> entire scanner bed" (instead of view -> item on scanner
bed); and then
2) set a selection border (otherwise it will mostly straighten and/or
crop when saving the scan even if the preview isn't straightened!)

Now, scanning without the lid, the preview (of my deliberately rotated
object) is straightened and cropped in spite of "view -> entire scanner
bed". Aarrgh!!
Apparently now it just doesn't *recognize* what the "scanner bed" is so
the view setting doesn't make any difference and I don't even get the
chance to set a selection border.

I found a solution, too.

Reasoning that apparently it needed a bit of "whiteness" to distinguish
the edge of the scanner bed, I cut a narrow strip of white paper (narrow
enough to still leave some room around the object) and put this along
one side of the glass plate. Bingo! I now get to see the whole glass
plate again and can proceed to set the selection border...


(What the heck am I doing here? Second-guessing a piece of dumb
software? Actually the software is cleverly designed to make things easy
for most people most of the time and its assumptions are reasonable -
the only problem is a missing "expert mode". Oh well, I learned
something again today!)

 

[Don]

I'll try that next. I've been experimenting with various cloths (scarves
and pieces of clothing!).

Cloth is fun to scan all by itself! ;o) When I was doing my photos
there was an album with a very intricate cloth cover. So, just for the
fun of it, I scanned the cover to include with the photographs. Like I
mentioned before I'm "digitizing my life" so "archiving" the album
cover as well as what's inside seemed like a thing to do.

Ah, yes, I had already noticed there seemed to be a few specks. Now
where was that thread about cleaning the glass plate? I'll want to clean
the inside of the cover, too, since when I do use the shadows the
background should be clean as well. (I'm pretty good with the clone
brush but anyway. ;-))

Me too, but it is time consuming. I'm fighting with "pepper spots"
right now which ICE can't fix anyway. So I turned ICE off completely.

The idea is to get back to it and do all that later (manually)...
Right now, I just want to "freeze" everything into the digital domain
where things don't deteriorate any more. I can then put all the
originals in storage and work on the images at my leisure later. Of
course, life will most certainly get in the way, and I will probably
never finish it - but it keeps me off the streets, as they say... ;o)

But correcting in individual channels is where I'm struggling now - the
scanner seems to always have a slight reddish cast. I'm working on a
(Chinese) bag of what is misleadingly called "American fried beans"
(actually something like "wasabi peas" if that means anything to you).

No, what are they? Is it a snack (like roasted corn or chickpeas) or
is it a staple food? As a vegetarian, I'm always curios about these
things. ;o)

Speaking of Chinese, the above album which cloth cover I scanned was
Chinese made!

The peas are bright green but somehow they come out brownish - and I'm
damned if I can correct it to the appropriate color without having a
negative influence on all the other colors. Trying histograms as well
as curves for individual channels (and combinations) but I just cannot
get it completely right (albeit a lot better than the actual scan).

One good way to get rid of casts is to change the color space. Lab is
usually recommended for removing a cast.

Also, it appears you need a complex curve with multiple points.
Normally a cast is removed by setting the gray point, but if the cast
is non-linear (like my infamous Kodachromes) then you need several
points because the cast is different in different areas.

Best explained with an example. Find a neutral area and set the
regular gray point as usual (at 128 with the 3 channels going left or
right as needed). Then do the same for the areas around 64 and 192.
This is tricky because you need to find neutral areas in the image
around these values. However, once you do that you'll end up with a
curve with three points each slightly different so the cast will be
gone without disturbing the rest of the image. Of course, you can use
fewer or more points at any place in the histogram as needed.

I don't print anyway, so pixels are what counts (and "inches" are
utterly meaningless). But then size also causes more pixels - but since
there's no rescan involved my guess is that's purely interpolation
(which can help but doesn't necessarily).

I don't print either.

But if the setting causes more pixels without a rescan then it does
sound like interpolation. In that case, as I mentioned before, I would
use native resolution and interpolate later in an image editor.

Ah, finally - you had me worried for a moment... :^)

;o)

Don.

 

[Don]

Oh, what fun! I tried that and ran smack into a new problem (or rather,
recurring problem that I tought I had a good workaround for):

The HP scanner software for this machine *automatically* attempts to
straighten a scan - using a really inferior technique instead of (for
instance) PSP's much superior resampling when rotating an image.

A-ha! The software is "doing you a favor"! That's when I shriek in
horror and ask: How do I turn this "fantastic" feature off? ;o)

The workaround I found was to
1) set "view -> entire scanner bed" (instead of view -> item on scanner
bed); and then
2) set a selection border (otherwise it will mostly straighten and/or
crop when saving the scan even if the preview isn't straightened!)

Now, scanning without the lid, the preview (of my deliberately rotated
object) is straightened and cropped in spite of "view -> entire scanner
bed". Aarrgh!!
Apparently now it just doesn't *recognize* what the "scanner bed" is so
the view setting doesn't make any difference and I don't even get the
chance to set a selection border.

The sad thing is the "programmer" (and I use the term very loosely
here) is probably very proud of this "feature" and how "smart" he was
to think of this. :-/

I found a solution, too.

That'll teach him! ;o)

Reasoning that apparently it needed a bit of "whiteness" to distinguish
the edge of the scanner bed, I cut a narrow strip of white paper (narrow
enough to still leave some room around the object) and put this along
one side of the glass plate. Bingo! I now get to see the whole glass
plate again and can proceed to set the selection border...

Good for you! Well done!

(What the heck am I doing here? Second-guessing a piece of dumb
software? Actually the software is cleverly designed to make things easy
for most people most of the time and its assumptions are reasonable -
the only problem is a missing "expert mode".

Was the "programmer" (again, using the term very loosely) named "Ed"
by any chance? ;o)

Don.