HOME
TOPICS
ABOUT ME
MAIL

 
  technofile
Al Fasoldt's reviews and commentaries, continuously available online since 1983

T e c h n o f i l e
Your scanner is telling you a little white lie about resolution


Feb. 23, 2003


By Al Fasoldt
Copyright © 2003, Al Fasoldt
Copyright © 2003, The Post-Standard

   Life is full of little white lies. Lumber that seems to be 2X4 isn't two inches by four inches -- it measures a lot less in both directions -- and your computer monitor isn't really as wide as you think. (The claimed size is based on a diagonal measurement, not from a measurement of the actual width.)
   Alas, scanners suffer from the same fanciful interpretation of numbers.
   One way to measure the quality of a scanner is to find out how much detail it can pick up. By "detail," I'm referring to the difference between noticing a tiny portion of a scene and ignoring it because it's too small. This is called "resolution." The term comes from "resolve," meaning, in this case, an ability to clearly show separate objects that are close together. (A lens for a camera or telescope that has good resolution is said to have good "resolving power.")
   Scanner resolution is rated by the number of closely spaced lines that can be "resolved" (seen clearly by the scanner) in a standard test page. For example, if a scanner can just barely show 300 lines spaced evenly across a 1-inch space, it is rated at a maximum resolution of 300 lines per inch. In the perverse world of computer jargon, this usually is called "dots per inch," or dpi, instead of "lines per inch," or lpi.
   All good scanners can reach a resolution of 300 dpi. Many can go higher, but most of us don't need more resolution than that. Buying an expensive scanner because it has a claimed resolution of a gazillion dots per inch doesn't make sense if you're scanning photos to put on Web pages or creating pictures that you can put on CDs to mail to grandma. For those uses, 300 dpi is more than enough.
   But the race for resolution won't stop, whether it's necessary or not. Scanner manufacturers seem particularly tempted to make their numbers go higher and higher every model year. Alas, that leads them down the lane of little white lies.
   They give you fake information by using an old trick. Here's how it works, using an example from ordinary life.
   Suppose the local Boy Scout troop recruited you to count the number of marchers in the annual Independence Day parade. You might pall at the thought of counting every drummer and high stepper for an hour or two. You might get clever and do some math.
   You look at the parade and see that most of the marchers are coming down the street in five rows. You decide to count all the marchers in the first row, the one nearest to you -- the one easiest to count -- and multiply by five.
   Cool, right? It's fast and simple, and you don't have to do a lot of work.
   But it's not so cool and not so right. Anybody who's ever watched a parade knows that the tuba player on the left might not be in step with the drummer on the right. Counting one row and assuming that all the rows are perfectly lined up is almost guaranteed to give you lousy math and an inaccurate count. If you're that lazy, you surely wouldn't make a good parade counter.
   Once they get past their native counting abilities, scanners usually do the same sort of thing when they're asked to scan more than they're capable of. They examine one row of marchers -- one row of "dots," that is -- and skip a row or two before examining another row. The ones in between are ignored during the first part of the scan, which I call the optical portion.
   The interpolation portion of the scan comes next. All those rows that were ignored suddenly show up in the scanner's portrayal of what it saw, as if by magic. But the magic is just a trick of math.
   Let's suppose your scanner saw a pattern of mostly light-colored dots at the beginning of Row 1, followed by a pattern of mostly dark-colored dots at the other end of Row 1. Suppose it saw a similar pattern in the next row it was able to scan -- Row 3, let's say.
   Keep in mind that it never scanned Row 2. It couldn't scan that row because it doesn't have enough actual resolution to scan it. It's just not good enough to make out that second row.
   But instead of giving up on that row, your scanner merrily offers to view it mentally, so to speak. It makes an educated guess and creates Row 2 out of thin air, making it more or less the same as Rows 1 and 3.
   And it does this for Row 4, using the actual optical view of Rows 3 and 5, as well as for Row 6, using -- well, you get the point. Every other row is guesswork.
   You might think that this would cause an outcry. But scanner owners seldom realize this is happening, and most computer users seem unwilling to examine the way things work anyway. They apparently think computers and peripherals such as scanners are too hard to figure out, and they blame themselves, not their hardware, when things don't make sense.
   But this is downright nonsense. You're the one who's smart. Your scanner is the dumb one. Don't let it get away with pretending to be better than it is.
   If your scanner, like most these days, has an optical (or real) resolution of 300 dpi, it probably can pretend it's scanning at a much higher resolution through interpolation. When it fills in for the dots it can't make out in both the horizontal and vertical directions -- a very common practice -- your 300 dpi scanner can perform what seem to be 1200 dpi scans (300 doubled horizontally to 600, then doubled vertically to 1200). They're not really 1200 dpi scans, of course. They're mostly faked. (Out of every four dots, three of them are made up.)
   Scanner manufacturers hope you don't mind when they fuzz up the meaning of "resolution," but you should take what they say -- and what the new-on-the-job store clerks say, also -- with a lot of skepticism.
   The only number that counts is the one that expresses the scanner's optical resolution. If you don't see it mentioned, you can assume that it's 300 dpi, regardless of the resolution listed on the box. If the scanner has two numbers, such as 1200X4800, the optical resolution is no higher than the lower number -- but it could be less than that number, of course.
   And when you scan, don't let high numbers fool you. If you can't find out what the real resolution of your scanner is, rely on your own good judgment. Make some tests and save them for comparisons. I'll tell you how to to this, without spending a cent, in a future column.