Scanning and Image Processing
The following describes the procedure I use when scanning and processing images of coins,
including the ones in my gallery at
http://www.lotn.org/~calkinsc/coins. The following is not intended to imply
this is the one and only method to produce images of coins, but it is the one that I have found
to produce the best images - it works for me, but your milage may vary. The details are rather
hardware and software dependent, but should be able to be generalized to other equipment and
image processing packages. If you have any questions or comments, please feel free to ask me
and direct them to calkinsc@lotn.org.
THE SETUP
THE PROCEDURE
THE END
Addendum - The QX3
The equipment that I use to produce my images is as follows.
The most important critera is to use a scanner of reasonable quality. What is "reasonable"?
As coins are generally small relative to photographs and other typical items that are scanned,
the scanner should have a reasonably high optical scanning resolution - this is normally
measured in dots per inch, and especially if you wish to scan particularly tiny coins, 1200dpi
or better is recommended. Scanners are often advertised with both an optical and an
interpolated or other enhanced resolution. Interpolated pixels are essentially "faked" - an
image scanned at a resolution higher than the optical resolution will produce a larger image,
but the image won't contain any more information than one at the optical resolution. The larger
image may actually look streaky or have other problems depending on the exact conditions of the
scan.
The scanner I use is a UMAX Astra 1220S, with a 1200x600 optical scanning resoultion. I had
previously used a UMAX Vista S6E with a 600x300 optical resolution, but upgraded to the Astra
as the smallest coins I wished to scan, such as Maundy pennies, looked "muddy" as the resolution
wasn't high enough to see much detail. The "S" in 1220S indicates the scanner has a SCSI
interface. Scanners with a parallel interface would work just as well, but a SCSI scanner should
operate somewhat faster - if one does a lot of scanning, this may make a difference in the long
run. A SCSI scanner is typically a bit more expensive than a parallel scanner, however.
There are other ways to obtain a coin image - digital or film cameras come to mind - but I find
a scanner is the most convenient. A camera allows greater control over lighting, but can require
extra equipment (e.g. light or camera stands, special lenses) or a longer setup to produce consistent
images. Digital cameras do not yet support the resolution of flatbed scanners, and film cameras
require the film to be developed and a picture created before that picture can be scanned. If the
equipment is of quality this may overall produce a better image than the scanner itself could,
but is much more time consuming to take the image itself, let alone the delays of film processing.
Scanned images rarely are perfect directly from the scanner, so some amount of image manipulation
needs to be done. Having a machine which is capable of doing it relatively quickly helps reduce
the "frustration factor" and let you finish up that much more quickly. My machine is a
dual Pentium II-450 with 256 MB of RAM and a lot of free hard disk space. Memory is more important
than CPU speed, especially as scanned images tend to be large - if the image can be manipulated
entirely in memory, a slower processor will work as well as or better than a fast processor on
a machine that has to continuously swap to disk. As soon as virtual memory is needed, performance
will drop considerably. Also because scanned images tend to be large, a lot of working hard disk
space is useful as the image is processed - the final image can be reduced in size, but as it is
being manipulated several tens to a hundred megabyte or so of free disk space may be necessary
to accomodate it (or other temporary files generated by the image processing software).
There are a number of image processing programs available, but I use the full version of
PhotoShop 5.5 - I find it has the most features and can produce the nicest images relative to what
I have used in the past. The full version of PhotoShop is rather expensive, however, which
may preclude some from using it. The light version of PhotoShop, PhotoShop LE, often comes
with scanners, but is missing certain features that I rely on when processing images. PhotoShop
is also a memory and disk hog when processing images - it maintains several copies of each image
for "undo" and other internal uses, so even if an image is of reasonable size, one might need
4 or 5 times its size worth of RAM or hard disk to process it. On certain images that I've
processed, even 256MB of RAM hasn't been sufficient to keep PhotoShop from paging. Generally,
all is well, however.
The scanner itself uses the VistaScan software, version 3.51 - by installing VistaScan, PhotoShop
can use this from its Import->Twain 32... menu to perform the scan within the context of
PhotoShop itself, rather than outside as the standalone scanner software allows.
As given above, this isn't the one and only way to create coin scans, but its the one that I
use and have had good luck with. Some of the steps may seem obvious, but I include them
anyway in some detail as even some of the obvious steps have pitfalls.
- Clean the scan bed.
Fingerprints or dust particles on the scan bed can appear in the scanned image, so cleaning
the scan bed before a scan is important. Although in later image processing any stray dust
particles and such can be eliminated, the fewer there are to begin with the better. It is most
important when scanning toned copper, or other dark coins. Scanners tend to auto-adjust based
on what is being scanned, so a dark coin will cause the final image to be considerably lighter -
even the tiniest dust particles will show up in the image as white dots on the coin surface.
As silver coins tend to be lighter this is less of an issue - the dust particles aren't as
accentuated - it still helps to have a clean scan bed.
- Align the coin on the scan bed.
This is actually more difficult than it sounds. If placing the obverse and reverse of a coin
next to each other in the final image is the goal, it is good if the light hits both the obverse
and the reverse at the same angle - the resulting image looks odd if shadows or highlights are
inconsistent between the faces. For instance, place the coin as close to "right side up",
relative to the face on the scanbed, as possible.
Keeping that in mind, however, light from a scanner often arrives at an angle, and as such
different features on a coin will be highlighted depending on its orientation to the light.
Consider:
This is a Durotriges stater - its the same coin in both images, just placed on the scanbed
differently as shown by the black box around each scan. Note in particular the legs of the
disjointed horse - in the scan on the left they are almost invisible, but in the scan on the
right they are prominent. As you can imagine, I had to scan this coin several times before
I was able to have all of its features visible in the scan.
Sometimes this can work to one's advantage however. Consider:
This is a 500 Peseta coin of Spain - note the "hologram" in the upper right of the reverse.
Depending on how the light hits the coin, either a 98 (the year of issue) or a crowned
M appears. I was able to produce scans of each by rotating and scanning the coin several
times until I obtained scans where one or the other was most clear.
I've found that coins in flat holders (e.g. certified slabs) scan well enough, but if
the holder is even the slightest bit wavy (as 2x2s can be), its best to place the coin
directly on the scanbed. This can be a bit of a trick, especially when wearing cotton
gloves to handle a more valuable coin to scan, but the best image is obtained. Wavy
holders can produce distortions which are hard to correct such as:
- Do the scan itself.
When VistaScan opens, it allows the scanner settings to be modified as well as providing
a preview window of what the scan will look like. These settings only need to be set once -
they will remain in effect until explicitly changed. The resolution of the scan should
(ideally, subject to system performance) be as high as the optical resolution of the
scanner, but no higher. As my scanner is 1200x600dpi optical, I scan at 1200x1200dpi.
Resolution does matter, as shown here:
This is a scan of one of the characters on the bottom of the obverse of a 5 Chiao from
Meng Chiang (1930s Japanese puppet state in China). From the left to the right, the scans were performed on my
scanner at 1200x1200dpi, 600x600dpi, 300x300dpi and 150x150dpi, and resized so they all have
the same physical dimension. If you scan at a resolution higher than what the scanner's optical
resolution is, this essentially happens in the scanner itself (or the scanner software) - I
did the resizing in PhotoShop after the fact. Note that the first and the second image aren't
significantly different - as my scanner is really 1200x600, there is only twice as much
information in the first scan vs the second - the other scans each have 4 times as much
information content than the next. This character is about 65/1000 of an inch across - small,
but seen easily without the unaided eye. It requires at least a 600x600 scan to be clear,
however. In other words, scanning at the highest resolution the scanner can provide is a good
idea - the final image can always be reduced for presentation purposes.
The scan should be in RGB mode - I personally prefer color, rather than black and white
images of coins, as even on a silver coin a little bit of toning that shows in the
scan gives the image more character than would otherwise be seen. Filtering and descreening
should also be disabled - any filtering will be done later, and descreening is only needed
when a preprinted photograph (which itself is made of dots) is scanned. The scanner
auto-adjust checkbox should remain checked - its a good idea to let the scanner determine
its own brightness and contrast settings on a scan by scan basis.
Click on the preview button, and a quick scan will be made which shows where the coin
is oriented on the scanbed. The crop box can then be used to place a rectangle around
the coin itself to indicate where the scan should be made. The scanner auto-adjustment
adjusts based on what it sees in this rectangle - the less background it sees, the
closer the scan will be to what is appropriate for the coin. (darker coins will have
scans lightened, etc.) I personally scan coins with the scan lid open - this produces
a black background as there isn't anything to reflect the scanner light back to the
CCD. I find that dark copper coins in particular are more visible this way as the
scanner doesn't have to overcompensate in its adjustments.
When the preview is complete and the scan rectangle marked, do the scan itself.
When complete, it will appear in a PhotoShop window.
- Scan the other face.
Do the above, but the other side of the coin - PhotoShop will now have two open windows,
one for each face.
- Safety first.
Exit the scan dialog and save both images. Use TIFF, Targa, PSD or other format that
does not lose data (i.e. has lossy compression). JPEG, for instance, will throw away
some pixels and approximate others to reduce filesize - this can be done when the
processing is complete, but the raw images should remain identical to what the
scanner saw. By keeping the raw images separate from the rest of the processing,
the original images can always be used again if the processing produces an effect that
was not as desired - if the images are not saved at this point, the scans themselves
would have to be done again. Considering that I don't like to touch coins at the
best of times, the shorter the time they are out of protective holders the better.
A mention of scanning proof coins should be made here. Modern proofs, due to the
mirror fields reflect light very very well. This can cause additional offset
reflections (since the light from the scanner itself is offset) to be captured.
Suppose the coin is in a plastic capsule and is lying on the scan bed - a reflection
could be seen off of the interior face of the capsule as well as off of the scan bed.
This is part of a scan of an Eisenhower dollar that shows this effect:
Scanner-induced doubling can easily be seen in the fields of the various coin features. Several of
the brightest features have washed out the scanner and have created green streaks
as well. Making a proof coin look "proof" can take a lot of hand editing, although
the result can be impressive, as the similar fragment of a proof Susan B. Anthony shows:
Removal of the scan-induced features follows similar techniques as described in
Remove Dust. The idea is that the stronger the mirror, the blacker
the reflection will be. Perfect proofs have perfectly black mirrors, so painting (or
erasing) the affected areas with black corrects the image and restores the proof effect.
These manipulations, since so much human labor is involved, can make a proof coin take an hour
or two to completely process, rather than the 15 minutes of a circulation (or non mirror
proof) strike.
- Rotate one face.
A coin on a scanbed will almost never be perfectly "upright" and the resulting image
requires rotation. Choose one of the images, and select Image->Rotate Canvas->Arbitrary.
A small window will be displayed to allow a rotation amount in degrees to be entered,
as well as whether its a clockwise or counterclockwise rotation. Type in a value that
seems appropriate, and do the rotation. If the value selected was not correct, undo
the rotation (CTRL-Z) and try again - its better to undo the rotation and do it again
rather than rotating an image which has already been rotated - as with scanning at
a resolution higher than the optical resolution of the scanner causes pixels to be
interpolated (a.k.a. "faked") that also happens here when the rotation is not a
multiple of 90 degrees. Although the number of changed pixels may be slight, the
effect is minimized when only one (correct) rotation is made.
An easy way to test to see if a rotation is enough is to move one of PhotoShop's
tool palettes over the image. Coins often have horizontal or vertical features, e.g.
crosses, shields, straight lettering - or at least symmetries, e.g. a dot on the
left is exactly as far down from the top of the coin as a dot on the right. Moving
a palette in such a way that its top edge or left edge is coincident with the feature
will show if the feature is aligned correctly - the top (and bottom) edges of the
palettes are exactly horizontal, and the left (and right) edges of the palettes are
exactly vertical. For instance, the shield as shown here aligns with the
Channels palette (arbitrary choice) in both its vertical (center divider) and
horizontal (top border) features, so I know it is rotated correctly:
- Safety first.
When the rotation is satisfactory, save the image. You may wish to save it under
a different name to preserve the original scan, but that is up to you.
- Rotate the other face.
Do the same
for the other face, and save that as well.
It is a good idea to remove any part of the image which is not related to the coin - this
can reduce file size and otherwise improve the look of the final image. There are
a number of ways to do this using the various tools of PhotoShop.
By using the Marquee Tool.
PhotoShop allows selections to be made that are rectangular or elliptical. By selecting
the elliptical tool (click on the dashed square in the main palette and hold the mouse
button down - a popup menu will appear showing other marquee types - select the dashed
circle), selections can be made that are exactly the correct diameter for the coin.
Once the elliptical marquee is active, go to the Marquee Options tool palette - this
palette allows specific parameters to be set for the marquee - set the Style to be
Fixed Size and type in values (which must be equal to get a circle) for the Width
and Height. Return to the coin image, hold down the ALT key, and click with the mouse
in the center of the coin. Holding down ALT will place the marquee - the selection
circle - by its center. You can use the arrow keys to move the selection to fine tune
it to enclose the coin completely. If the diameter is too large or too small, press
CTRL-D to clear the selection, type in new numbers for Width and Height, and try again.
When the correct size is found, press CTRL-C to copy it to the clipboard, CTRL-N
to open a new window - PhotoShop defaults to setting the size of the new window to
match the size of what is on the clipboard - and click on OK to create a new empty
window. Press CTRL-V to paste the selection in the new window - the hatch pattern
indicates that there are no pixels outside the coin, and thus the background is
removed.
Note that while this works for milled coins, hammered coins or other oddly shaped
coins aren't circular. The above method is the fastest of the selection methods given
here, but isn't always the most appropriate choice. As shown here, a good selection
should touch the very edge of the coin, as with this milled shilling - spade coins,
however, are not round:
By the Lasso Tool.
Just under the Marquee Tool is the Lasso Tool. The normal lasso looks like a rope -
this is for freeform selections. Clicking the mouse button on it to display a menu
shows other lassos - the Polygonal Lasso allows one to connect points by straight lines
and is most useful. The background of a coin can then be "cut away" - the lasso selecting
a region around the edge of an irregularly shaped coin, and once the selection is closed,
CTRL-X cuts the enclosed region out. Alternatively the coin itself can be selected, and
the copy and paste to a new window method above can be used. It takes a bit longer to
select in this way, but its the easiest method for irregular coins.
By the Magic Wand.
To the right of the Lasso Tool is the Magic Wand. This tool selects pixels by color, and
may be used to select the background in one fell swoop instead of piece by piece as with
the Lasso Tool. It also has the ability to select colors similar (within a tolerance)
to the current color, so if the background differs slightly (e.g. black to not quite black)
it still may be able to select the full region. It may, however, select part of the
coin instead if the colors of the coin are close enough to the color clicked on. Areas
can be unselected by holding ALT and using the other selection tools, but often this
tool is more work than its worth unless the color to be selected is uniform. For instance,
if the Lasso Tool is used to cut out parts of the background, the backround may be
replaced instead by the PhotoShop background color, e.g. white. Once the background is
all white and if there isn't any white in the coin, the Magic Wand tool can be used
to select all of the white at once. Inverting the selection (via the Select) menu
then selects everthing that isn't already selected - swaps selecting the background for
selecting the coin instead. Again, cutting and pasting to a new window will create an
image which has the background removed.
Safety First.
As always, save the file. To maintain the empty background as empty, it must be
saved as a PhotoShop (.PSD) file. Do the same for the other face and save it
as well.
I like to have the obverse next to the reverse in a single image.
Get the size of one face.
Choose the image of one of the faces, press CTRL-A to select the whole image, and
CTRL-C to copy. If you had used the Marquee Tool to select both faces originally,
you can go on to the next step. If you didn't, go to Image->Image Size (or
File->New) and write down the width and height shown.
Set the Width of the other face.
Go to the image of the other face. Choose Image->Canvas Size. If both faces
are the same size, the dropdown next to Width can be changed to "Percent"
rather than "Inches" and 200% typed for the new size. If the faces are not
the same size, add the width written down of the other face to the one given
for this face.
Set the Height of the other face.
In that same dialog, the Height may need to be adjusted. Again, if the
faces are the same, no changes need to be made. If the height of the first
face (the one that you pressed CTRL-C on) is greater than the one here,
set the Height of this face to match the other face. If its equal or less,
leave it alone.
Shift the face to the correct position.
Next to the Width and Height settings are nine squares, and by default the center
square is pressed. The pressed square indicates where the current image will
be when the canvas size is changed. With the square in the center, the
face will remain in the center of the newly enlarged window - not what you
want. You want to select either the left square (if the face being resized
is the obverse) or the right square (if the face being resized is the reverse).
Resize the canvas.
Click on OK. The canvas will now resize and the current image shift where
the pressed square said it should go. The face should be up against either
the left or right side of the window, and a large empty space on the other side.
Paste the other face in.
If CTRL-V is pressed to paste the other face in, the image will be pasted
in the center of the open window - this may require extra effort to shift the
image into place, but there is an easier way.
Shrink the window horizontally so the amount that is visible is smaller than
the empty space (if, say, the image was 1000 pixels wide to contain 2 500
pixel faces, shrink the window to, say, 400 pixels - a value smaller than 500).
Scroll the window all the way to the edge so only the empty space is seen
an none of the face currently in the window. Be sure the scroll bar is as
far over as it can go - you must be right to the edge. Now use CTRL-V to
paste. The first face now falls exactly in the empty space, right next to
the face that was already there - by being all the way over to the edge and
having less area visible than the size of the face to paste, PhotoShop knows
to paste the image right up to the edge of the window, where you want it to go.
Safety First.
Once again, save the resulting image. The windows of the individual faces can
be closed as you won't be needing them again.
Of course, a corresponding procedure works if you wish to place the faces top to
bottom, or you wish to place a different image altogether (such as a magnification)
next to the coin. Reduce the problem to two images, increase the canvas size
of one by the appropriate dimensions, and paste the other in.
Scans often come out too dark, too light, or otherwise washed out. Through PhotoShop's
adjusting of the image levels, most of this can be compensated for.
- Enter Lab color space.
The full version of PhotoShop supports multiple color spaces. Up to now the process
has worked in RGB, representing "channels" of Red, Green and Blue. Each channel
represents the intensities of that color. For instance, a black pixel has a
RGB value of (0,0,0) - 0 Red, 0 Green, 0 Blue. A white pixel has a RGB value of
(255,255,255), the maximum values of 255 for Red, Green and Blue each. Dark blue
has a value of (0,0,128). RGB is useful to display images, but its better to
manipulate them for this procedure in Lab color space.
Lab mode still has three channels, but these channels are L, a and b. a and b
represent the colors of the image, and L the luminosity - essentially a single
value that represents the intensity of the color at each pixel. A value of 0 of
the luminosity of a pixel makes it black, a value of 255 makes it the most intense
color that the a and b channels represent. The luminosity
channel alone looks like about what the image itself would look like if it
was converted to greyscale (e.g. printed in a black and white newspaper).
Lab mode can be selected via Image->Mode->Lab Color. When composing the image in the
step above, PhotoShop 5.5 places each of the two faces on separate layers -
a layer is essentially a drawing surface that is overlayed on to the layers
below. If both faces have about the same intensity (if one face wasn't scanned
obviously brighter or darker than the other face) they can be processed together,
so when the dialog appears asking if you wish to merge the layers or not,
it is safe to say Merge. Both faces are now on the same layer.
- Select the luminosity channel.
If the Channels palette isn't already open, open it. The composite "Lab" channel
should be selected by default and the image should appear in color. Click on
the "Lightness" entry in that list. The image should appear to be grayscale as
you are now viewing the lightness channel alone.
- Level the channel.
Select Image->Adjust->Levels to display the histogram level dialog. Since you
are looking at the luminosity channel only, the shape shown in the dialog
represents the distribution of pixels at that intensity - each bar represents
the relative count of pixels at that particular value - 0 is on the left,
255 on the right. If the image was mostly black, the graph should be skewed
left, if mostly white, skewed right. For instance, suppose you are processing
a Civil War Token as shown here (its shown here in color, though you would
be looking at the luminosity channel alone - a magnification of the upper
right quadrant is also shown to present a closeup of the effect)
For the lightness channel of this image, the leveling dialog looks like this:
The idea of leveling is to "spread out" the distribution so it isn't so concentrated -
the triangular sliders can be moved to make this happen. The black triangle
has the effect of all values to the left of the triangle being set to 0.
The white triangle has the effect of all values to the right of the triangle being
set to 255. The grey triangle represents the midrange of the image - it should
be set somewhere between the black and white triangles, and generally near the
peak of the graph is a good place. Here is a setting:
If the triangles are too close together, the image will become accentuated in
strange ways - you want to keep the majority of pixels between the triangles,
and eliminate the regions of the graph where no pixels were counted.
Clicking on OK and applying these values gives this image:
The change is subtle, but it looks less washed out - the wreath and the lettering
appear bolder. If the leveling dialog is opened again, it shows:
Note that the histogram now goes from edge to edge instead of being predominately
in the middle.
The reason to perform leveling on the lightness channel alone is that if the color
channels were used, the colors would be skewed as well. Suppose the image to be
leveled consisted of a green bar next to a red bar. After the leveling, the
red bar would become much darker, as shown here:
The idea is to leave the color of the coin unchanged, but make it more "intense."
With the intensities now adjusted, a small amount of sharpening to the lightness
channel can be applied which will bring out even more detail. Scanners seem to
scan "soft" where edges aren't always as crisp as they should be - a small amount
of sharpening restores this crispness. Again, it should be done only to the
lightness channel to prevent the colors from skewing.
- Flatten the image.
In general, the sharpening filter should be able to be applied at any time. Under
Windows NT on my machine, however, unless I flatten the image first, NT may crash
with a Blue Screen of Death. Not always, but enough to make me avoid it altogether.
To flatten the image, open the Layers palette if it is not already open. Click on
the triangle in the upper right of this palette - choose Flatten Image from the
dropdown. The layer name will change to Background. This is also important later
for saving the image - most formats don't understand more layers than just a single
"background" layer.
- Sharpen.
If an image has been sharpened before in this session of PhotoShop, pressing CTRL-F
will execute the sharpening command on this image, and the step will be complete.
(Technically, it applies the last filter used, but this procedure only uses one
filter - the sharpening one). If you haven't sharpened an image before, choose
Filter->Sharpen->Unsharp Mask. A dialog will be displayed where sharpening
settings can be entered. For my scanner, a radius of 4.0 pixels, an amount of 40%
and a Threshold of 0 I find give a good result. Click on OK to sharpen the image
with those settings. The same Civil War Token as above, after sharpening, now looks
like:
The effect is again subtle, but on the lettering in particular some of the edge "fuzz"
has gone away and they look more crisp.
- Possibly sharpen again.
I normally find a single sharpen operation is enough, but sometimes a second sharpening
will bring out even more detail - just press CTRL-F if needed. If it looks too extreme,
press CTRL-Z to undo the sharpening.
Depending on how clean the scanbed was this step may be skipped, but small dust particles
can be removed from the resulting image.
Return to full color.
Click on the composite Lab channel in the Channels palette to return to full color,
rather than the grayscale of the lightness channel.
Rubber stamp away the dust.
Select the Rubber Stamp tool, the fourth tool in the left column of the main palette (it
looks like a hand stamp). This tool allows you to paint on one place of the image using
the colors of another place of the image. Look at the image and find a dust particle -
you can use + and - to magnify and reduce the image in the window to get a closer look
at one. Hold down ALT and click on the image on a place with a color similar to what would
be under the dust particle - for instance, on a copper coin that has a dust particle,
click on a place of the image that has the same tone of copper as would be under the
dust. Now release ALT, and click (and drag, if its a large particle) with the mouse
over the dust particle. PhotoShop will now paint using the color from where you clicked
when you were holding ALT. Note that it paints using the color from the relative distance
as the mouse currently is from where you clicked - it doesn't use the exact same color
all the time. This is often better as even the most perfect coins still have surfaces
that vary slightly - always painting with a single color, e.g. brown for a copper coin,
will replace the white dust particle with a brown smudge that doesn't blend in with the
coin surface. Repeat this procedure (ALT click, then paint) for any other dust particles
that you find.
WARNING WARNING WARNING WARNING
Note that this will also allow you to brush away any defects in the coin itself - the pixel
showing a dust particle is no different than a pixel showing a scratch, dent or hole in the
coin. If you are trying to create an "ideal" picture of a certain type of coin you may
wish to brush away blemishes and defects, but if you are trying to represent the coin
as it is, don't make any modifications that would change the surface itself. It would
be easy to, say, deceive a prospective buyer of a coin you are selling to represent it other
than it is. As an example, I have a damaged U.S. Seated Dollar - the top image is how the
coin actually is in real life. The bottom, however...
Note how the coin surface is a mottled grey - this is why painting with a single color
isn't helpful to remove dust particles - a swipe with a solid color would stand out
as much as the particle itself would - by painting with the mottled grey, however,
the image looks much more believable.
Finally, the resulting image can be saved as the procedure is complete.
Return to RGB color space.
Choose Image->Mode->RGB Color and the image is converted.
Its important to return to RGB mode as most file formats won't support any
color mode other than RGB.
Save the file as a JPEG.
I like JPEG as the final medium as the files are small yet the full 24 bit color is
maintained. JPEG images can also be viewed by all graphical web browsers nowadays, so there is
no danger of someone not being able to view the image. As JPEG is lossy, it should only
be used for the last step. The amount of loss is controlled by the "quality" factor
when saving the file. I use PhotoShop's "High" quality setting - in version 5.5, that
represents a value of 8 out of 12, in earlier versions it represented 6 out of 10. Lower
quality factors reduce file size but increase the blockiness of the image.
When saving the file I typically give the file a suffix matching the resolution
of the image - in this case, _1200 as it was a 1200 DPI scan. That Civil War Token could
then have the name, say, cwtoken3_1200.jpg. The resulting filesize, though much smaller
than the original TIFF (or PSD) scanned image, is still too large for the web, though I keep
the image in case I need the full version in the future.
Save other resolutions for convenient web access.
Choose Image->Image Size and select "Percent" as the units. Change the value to 50% and
click OK. The image is now 1/2 the size in each dimension, so 1/4th the size in area.
Save it again with the same prefix, but a _600 suffix. Resize, and save as _300.
Resize and save as _150. The _600 and _150 sizes are now suitable for my web site
as the full image, and thumbnail, respectively.
That's it. Its long to describe, but the steps themselves aren't complicated - a non proof coin
can be scanned, processed, and saved in about 15 minutes. This method is probably impractical
for coin collections that change rapidly (e.g. a dealer's stock), but are good for galleries of
static (or increasing) collections, or for providing representative images of types.
Although the majority of my images are taken and processed as described above, I have obtained
an Intel Play QX3 digial microscope for further images. Although technically a kids toy,
its surprisingly good for imaging small coin features such as doubling. This microscope has
three magnification settings - 10x, 60x, and 200x, and captures images in 352x240 by default.
60x I have found the most useful for most of the images I have made, though 200x is occasionally
useful - a Morgan dollar mintmark won't completely fit in the field of view at 200x though, so
the feature to image really does need to be tiny for 200x to be useful.
Although the microscope comes with its own software to capture images, it also ships with a
TWAIN driver - images can be captured via PhotoShop just as with a regular scanner. I thus
acquire and process images in a similar way as to what is described above.
