The following is a guest post by Barry Wheeler, Digital Projects Coordinator, Office of Strategic Initiatives
In part 1 of this blog series, we saw that manufacturers claimed “resolution” is based on the number of steps per inch a small motor moves the scanner assembly (the rows) and the number of tiny sensors per inch the manufacturer puts on the assembly (the columns.) But the International Standards Organization does not consider this the scanner’s resolution. The ISO defines the steps and sensors as the “sampling” rate because the sensors can only attempt to measure (sample) the brightness at each point. The ISO defines “resolution” as the actual result on the screen, not the number of sensors and steps that attempt to read each point.
We can think of the primary difference between measuring each point and actually resolving each point as “efficiency.” Some of these differences come about because the light may scatter and miss the sensor, the motor step may not be sufficiently precise, or the collected value may be inaccurate. Inside every scanner or camera, between the sensor and the screen is a small, highly specialized computer called a digital signal processor. This processor must work very hard to link a dot on the page to a dot on the screen. Unfortunately, sometimes a manufacturer attempts to compensate for scanner inefficiencies by over-processing the image data, introducing new problems.
The Library follows the ISO standards and scans a “target” such as this one in Figure 1.
Then computer software applies sophisticated ISO specified formulas to measure the actual resolution. Targets and technology are necessary in a production environment, but you can do the same using your own images and your own software – your brain – to evaluate the resolution of an image. The targets used by the Library include figures to visually evaluate a scan. I’ve enlarged the central region of our target in the images that follow. By showing you how we examine this region, I hope to help you learn how to evaluate the resolution of your own scans.
Any numbers are only approximate but the analysis will show you if the resolution is sufficient for your own needs. Figure 2 shows an almost perfect capture of the target area from a very high priced scanner operated at a very, very slow speed.
Figure 3 shows a capture of the same area from a high quality camera with an excellent lens.
Figure 4 shows the capture from an inexpensive scanner (before it was dissected for our earlier blog!).
While Figure 4 shows us an unacceptable scan, Figure 5 shows a much improved scan from the same scanner! By experimenting with the scanner software controls – and carefully examining the results – a much improved scan was produced.
Now we can answer the question, “What resolution should I use?”
Assuming you will look at, or print, the image at the same size as the original (enlargements and film scanning will require another blog post), begin with a scan resolution of 300 dpi or 400 dpi. Choose a document to scan with some areas of fine lines and detail. Examples might be eyelashes, eyebrows, hair, small tree branches, perhaps detail in clothes, or many small windows in a building. Examine the resulting scan carefully. Look for clean solid lines, distortion, poor contrast between lines, false color, edge and halo artifacts around text. Then change the resolution, or the contrast, or the “sharpening” – change any one control and run another scan. Repeat several times. With practice, pick the scan you find best.
That’s it! On to Part 3.
Comments (6)
Very interesting post, thank you for all the illustrations. I’d like to know what settings you were able to adjust on the scanner to get the second, improved, image. I realize these will vary from scanner to scanner, but knowing what you tweaked would give me an idea of what to look for on my scanner. Also, the target you scanned, is that available somewhere? Thanks for all the detailed information, I find your posts very useful and share them with my photography and scrapbooking friends, many of whom deal with old images from their genealogy research.
If I am right, you switched from B&W to grayscale, correct?
Scratch that. Looked closely at the image. So, was blur filtering the difference?
Thank you for your kind comment. To improve the scan quality I bypassed the one-button autoscan feature and went to the options fields. I systematically varied the options and made over 20 test scans. I found setting the scan resolution to 400 dpi produced better results than the 300 dpi, the 600 dpi, or the 1200 dpi settings. Almost all tests were better when I set the “sharpening” control to “none.” I found the other controls had less effect. But your results may vary.
I used a small portion of a commercially available target designed to evaluate many different scanner characteristics both visually and with additional software. But there are several free or inexpensive resolution targets available – they are sometimes included with a specialized scanner control software program. You can look for the WG-18 resolution target, the USAF 1951 resolution target, or the full-sized ISO 12233 target. http://www.imaging-resources.com provides a thumbnail image and description of each target which you can find through the site search. And the website “Preserving History” at http://archivehistory.jeksite.org provides a detailed explanation of how to use the USAF 1951 target in Appendix C. A Google search will turn up others helpful resources as well.
Let me add – the Library begins the digitization process by scanning to a “master” file. We set the scanner or camera to produce the most basic image possible with no sharpening, no increase contrast, no vivid or color enhancement. We want as little digital processing as possible in our master file. We save our master files in the .TIF file format. These files are big, but they contain all the detail the scanner was able to capture. We try to avoid the problems that scanners can create using the scanner’s limited processing power and simple processing methods. Later we create a “derivative” file for printing or web display.
Home and small institutions may want to simplify this process. Experimenting and looking at the results carefully can help determine how to operate the scanner for best results. Thus, for this particular scanner, 400 dpi scans were visually better than 600 dpi or 1200dpi scans. That’s an example of asking too much of this particular scanner’s limited power.
Then save the files in the highest quality jpeg format possible. It is difficult to create a single scan for all purposes. For instance, many files will be too big to email, but I don’t want to create a poor quality copy for email and then have to use that file to print an 8″ x 10″ for framing.
For more details, email me a [email protected].
For a test document to scan, I find Victorian engravings (in the original printed edition, not a reprint) most suitable. They have fine black lines and very small dots, and also very small white dots in the darker areas.