|
|
|
This document discusses in detail many aspects of creating digital images of coins and similar numismatic items with a scanner. Important technical factors are explained for those who may need or like to better understand the trade-offs and other issues involved in the process. Next, suggestions on equipment and software for scanning coins are provided. Finally, an easy to follow procedure gives step by step instructions for creating scanned coin images suitable for the world wide web.
NOTE: The original version of this document was written in early 1997. Digital cameras have advanced considerably since then. If you're deciding between purchasing a scanner or a digital camera, the camera may now be the better alternative, especially since it can be used for other things that a scanner cannot. Image "post-processing" - described in the Step by Step Procedure below - is similar.
Scanning Coins vs. Scanning Coin Photographs
Image Resolution
Monitor Resolution
Magnification, Storage and Download Time Considerations
Thumbnails
Image Formats
Choosing Hardware and Software
Scanned images of coins can be used for many purposes, not the least of which is to "spruce up" any numismatic web site. I began scanning coins and including the results on the web in 1994. After numerous requests for tips on scanning coins, I created this page to make detailed information available to everyone.
This discussion is oriented towards producing coin images to be displayed on computer monitors, such as on web sites. Many points are also applicable when the intended result is hardcopy.
Since the goal is not just to create coin images but high quality images that enhance a web site, let's first look at some of the major technical factors that come into play.
Is it preferable to scan coins directly or to photograph them and scan prints of the film?
The answer depends on your objective. Film offers higher resolution than scanners currently available in the local computer store. When the quality of the image is the overriding objective, photographing the coin and scanning the photograph usually produces the best results, provided that high quality equipment, film, paper and processing, as well as lighting suitable for coin photography are utilized. Scanning a photograph is generally advisable for high magnification images of coin details (for example, of die varieties) and may be necessary for very high relief or dark coins.
Placing the coin itself on a flatbed scanner is considerably faster and less expensive than making and scanning a photograph. This technique can produce images of up to 70 times actual size (see next two sections) suitable for most purposes, including personal and business oriented web pages. It usually works even with coins in slabs or other holders (for best results, remove coins from holders such as flips and 2x2s).
A digital camera may now provide comparable or superior results, provided it has sufficiently high resolution and a macro capability. Adapters for video cameras that digitize images are also available and may produce satisfactory results. Details on those devices are outside the scope of this document.
Prior to 2001, all coin images on the Telesphere Numismatics web site were produced by scanning coins directly on a flatbed scanner. We have since begun using a digital camera.
When an object is scanned, a digital representation is created. A matrix of discrete "picture elements" (pixels) representing individual points on the surface form a picture of the original object.
The image resolution or scanning resolution is the number of samples per unit of distance. The higher the scanning resolution, the finer the details that will be present in the resulting image. Of course, as resolution increases, the size of the file needed to store it increases, too.
The optimum resolutions to scan and display coins depend on factors considered below. In most cases, the image resolution will be between 72 and 600 pixels per inch, also known as "dots per inch" or "dpi" (divide dpi by 2.54 to get dots per centimeter).
Because an image has two dimensions - a width and a height - its size (number of pixels) increases proportionally to the square of the scanning resolution. When you double the scanning resolution, say from 150 to 300 dpi, the result is four times as many pixels.
Computer monitors display many thousands of pixels. The monitor resolution (or, more generally, device resolution) is a measure of how closely spaced the pixels are displayed. Monitor resolutions vary from one computer to another and are even configurable on some machines. In many cases, monitor resolution is at or near 72 dpi.
Because monitor resolution varies from one computer to another, the physical size of a given image also varies between the machines. An image that is 72 pixels wide and 72 pixels high will be displayed in a one inch by one inch square when the monitor resolution is 72 dpi. On another computer with a monitor resolution of 100 dpi, the same image will appear smaller, because the pixels are spaced closer together.
The very important point here is that the width and height of an image may be substantially different when viewed on a different computer or when printed. Some image processing programs report size in units such as inches or centimeters. In my opinion, that confuses the issue, because it's based on an assumption about device resolution which is invalid when images are shared over computer networks. It's more useful to know the width and height of an image in pixels, since they are device independent. To determine the physical size for any specific device, divide the number of pixels by the device resolution.
Scanning at higher resolution captures finer details in a coin. On the other hand, the resulting image occupies more disk space and takes more time to transfer over a computer network, such as when downloaded from a web site.
The longer it takes to download a web document, the greater the chance that the viewer will decide to interrupt the transfer and go to another site. My rule of thumb is that a web page should download over a 28.8K modem in under 30 seconds (some connections are slower and therefore download time is longer). That means the total size of all files used in the page should be limited to about 100 kilobytes.
Ideally, image magnification is high enough for the design details of the coin to be clear and small enough that transfer time is reasonable.
A
single high resolution coin image can be a few hundred kilobytes to
over
a megabyte. A useful technique for making large images available is to
include
a much smaller "thumbnail" version of the image in a document
and link it to a higher resolution image. Download time for the main
document
does not become excessive, and anyone interested can see the enlarged
version
by clicking on the thumbnail. The gold coin pictured here is an example.
After scanning a coin, you will save the image in a file.There are dozens of digital image file formats and a lot of software for converting from one format to another. Images used on the web should be in a format that browsers will display. The only two formats which virtually all web browsers are currently capable of displaying are:
GIF generally gives better results for digitally created art work, and JPEG generally gives better results for photographs. Either format is suitable for most coin scans. JPEG images can frequently be stored in somewhat less disk space. GIF has the advantage that an image can have a transparent background. One color can be designated as transparent - in our case, the color outside the perimeter of the coin would be selected. Whatever lies underneath transparent portions of the image (e.g. the background of a web page) will be displayed by software that supports transparency, including most browsers. In my opinion, transparency tilts the balance in favor of GIF for most coin images on the web.
If you don't already have a scanner, you may be wondering which ones are suitable for scanning coins (and anything else you expect to scan). You'll also need software to drive the scanner and to manipulate and save the images you create.
Handheld scanners are not suitable for scanning coins. A flatbed scanner is mandatory. You will need a model with a communications interface compatible with your computer. PC and Mac versions of many popular models are available.
A scanner that uses a Charge-Coupled Device (CCD) sensor is preferable to one that uses a Contact Image Sensor (CIS). CCD scanners are better able to capture objects that are not directly on the scanner surface, such as coins in slabs.
Many scanners come with software which attempts to produce higher resolutions images than the device can actually "see" by interpolating between pixels. This "enhanced" resolution is often cited in advertising for the product. Ignore it! The only figure that matters is the optical resolution of the scanner. No additional details are captured at resolutions higher than the optical resolution.
For professional use, I recommend a scanner with an optical resolution of at least 600 dpi. At 600 dpi, details of die varieties with significant "spread" can be captured. For personal use, 300 dpi may be satisfactory. Images on the web will often be at a lower resolution anyway. Flatbed scanners with 300 dpi optical resolution are available for under US$100, and 600 dpi optical resolution scanners are also relatively inexpensive.
Scanners are often bundled with an image processing program, and other software can be purchased separately. Among the most popular packages are Corel PhotoPaint and Adobe Photoshop. Keep in mind that you will probably want to use the software for a lot more than just driving the scanner (e.g. to annotate images or to create special effects). Photoshop is a great all around image processing program, but less expensive software will suffice if you just want to create scanned coin images.
All the coin images included on the Telesphere Numismatics web site prior to 2001 were scanned on an HP Scanjet 4C and subsequently manipulated as described below with Adobe Photoshop on an Apple computer. Other choices can also provide satisfactory results.
The procedure below includes instructions specific to this setup. Details will differ somewhat for other hardware and software.
Once your scanner and image processing software are installed, we're ready to scan! The process may appear to be long and complicated. That's only because I've described it in great detail, so that most anyone will be able to follow along. After a little practice, you should be able to scan, process and save a "web-ready" image of one side of a coin in 5 to 10 minutes.
Some of the instructions below are specific to my setup, but the process will be similar with others. Check the manuals for your scanner and software when you don't find options in the places indicated below.
Note: steps 16 and 17 were applied to the intermediate images shown at steps 9, 11 and 12, so that they would be visible in a web document.
1. Getting Started. Turn on the
scanner
and start up your image processing program. Place one or more coins on
the
scanner surface. A soft flip will help protect the coin from damage
when
the lid is closed. The Scanjet 4C gives the effect of a light source
located
at its front. Rotate the coin so that the light hits the coin devices
from
a complementary direction. Avoid dragging coins or other objects on
the
scanner surface!
2. Preview. Direct the image processing program to import an
image
from the scanner (in Photoshop 4.0, File->Import->Twain Acquire).
The software may launch another program that drives the scanner (for
the
Scanjet 4C, the DeskScan II program that came with it). Start a
"Preview"
operation, if it isn't launched automatically. A preview is a low
quality
image of what's on the scanner surface.
3. Image Type. Scanner drivers often
have
multiple image types, such as black and white drawing and color photo,
and
may guess which one best applies to what's on the scanner. "Millions
of Colors" or "Sharp Millions of Colors" work best for coins.
If necessary, override the "Type" accordingly.
4. Set the Scanner Resolution. You can and will tweak the image
after
the actual scan. First, you must get the raw working material, and the
most
important parameter to set at this point is its resolution. In general,
a final resolution of 150 dpi is satisfactory for larger coins, and 300
dpi is advisable for smaller ones. Scanning at higher than the final
resolution
enables a higher magnification image to be made later without
rescanning
the coin (reducing the resolution is covered in Step 11). A resolution
of
at least 600 dpi is advisable for RPMs, die doubling and other fine
details,
although going above the optical resolution of the scanner buys you
little
if any improvement in detail.
The interface to set the scanner resolution is effectively hidden by the DeskScan II interface. Click "Custom" and select "Print Path" from the pulldown menu. Enter the desired scanning resolutions in the Horizontal and Vertical Resolution boxes. Give the configuration a name, then click on the "Add" button followed by the "OK" button. You'll only need to do this once for each scan resolution. Later, you will be able to choose settings from the Path pulldown menu.
Change the "Path" setting to get the
desired scanning resolution, if necessary.
5. Zoom on the Area of Interest. Click on the preview image
outside
the rectangular selection area created during the preview to remove the
rectangle. Drag a new rectangle encompassing the actual area of
interest
(the rectangle need not be right at the borders of the coin, as excess
space
will be cropped later). Click "Zoom" in the software window to
get a close up of the area selected.
6. Adjust Brightness and Contrast. Re-select the area of
interest
again, if necessary. Note the Brightness and Contrast settings, in case
you want to return to them. Click the "yin/yang" button and the
Brightness and Contrast will be automatically adjusted for the
selection
area. Manually change the values, if not satisfied with the automatic
levels.
You'll be able to fine tune these parameters later, but any unusually
dark
areas should be removed now by increasing the brightness and/or
reducing
the contrast.
7. Import the Image. Click the "Final" button (or equivalent)
to scan the image into your image processing program.
8. Save the Raw Image. Save a copy of the image in the
software's
"native format" before doing much to it. If you later decide to
do things differently, you can open the saved image rather than
scanning
the coin again.
9. Cropping and
Rotating. Crop excessive
space around the coin. Rotate the image to the desired orientation, as
necessary.
In Photoshop 4.0, select "Rotate Canvas" from the "Image"
pulldown menu. An "Arbitrary" rotation angle is usually necessary.
In fact, you will usually need to rotate, undo the operation, and redo
it
a few times to get the right angle. Save the image again (this is the
last
time overwriting the originally saved copy is advisable).
Grey parts of the background in the adjacent image are from the original scan. White parts at the corners are the result of new pixels added to the image during rotation.
10. Color and Brightness.
Adjust the color and brightness settings, if desired (I usually don't).
11. Image
Resolution. The remaining steps
involve preparing your image for display, which is assumed here to be
on
the web. Reduce the image resolution to the final value, if not the
same
as the scanned resolution. For inclusion on a web page, 150 dpi is
appropriate,
except for smaller coins which may require 300 dpi for legends to be
clear.
In Photoshop 4.0, select "Image Size" from the Image pulldown
menu, change the "Resolution" to the desired value and click on
the "OK" button.
12. Sharpen the Image.
Details in the reduced size image can be made clearer by performing a
Sharpen
operation. For Photoshop 4.0, select "Sharpen" from the "Filters"
pulldown menu. The regular "Sharpen" option in the submenu that
appears seems to work best most of the time.
At this point we could jump to
step
16 and save a usable image. If you want to include both sides in one
file,
continue with step 13.
13. Scan the Other Side.
Repeat steps 2-12 for the other side of the coin.
14. Open a New Window. Create a new frame (File->New) large
enough
for both sides of the coin. Make sure the color mode is RGB, not
indexed.
Choose a background color that's not in the coin images (white usually
works).
15. Copy Images to the New Window. Next, we want to copy each
side
of the coin, minus the "noise" outside it to the new window. If
the coin is disk shaped, drag an elliptical marquee around it (easier
if
you first crop virtually all excess space on the left and top sides).
The
goal is to exclude everything outside the coin without losing any of
the
coin itself. Copy the selected area to your computer's clipboard
(Edit->Copy).
Drag the marquee into the new frame. Move it to one side of the frame
or
the other. Copy the contents of the clipboard into the new window
(Edit->Paste).
Repeat with the image of the other side of the coin. For a coin that's
not
disk shaped, this step will not be so easy, and you may have to live
with
some excess background from the scan.
16. Convert to Indexed Color. When the final image uses an
"RGB"
format such as JPEG, proceed to step 17. For GIF images, the image type
must be converted from RGB to Indexed Color. Because no more than 256
colors
can be displayed simultaneously in most web browsers - and that's for
everything
in it, not just one image - use no more than 128 (7 bit) colors. In
most
cases 32 or 64 colors (5 or 6 bits) are enough and result in a smaller
file.
If you're prompted about flattening layers, do it.
17. Save the Image. Select "Save As" from the File pulldown
menu. Set the output format to JPEG or GIF, as desired. Enter a new
filename
and click the "OK" button.
18. Make the Image Background Transparent (optional). Many
image
editing and web authoring tools enable one color of a GIF image to be
specified
as transparent. I use Adobe PageMill for this purpose.
Congratulations, you now have a high quality digital coin image!
