
How
Scanners Work![]()
![]()


A scanner serves as the eyes of a personal computer. Containing thousands of cells that work like the compound eye of a bee, it can see photographs, capture art work, and read hard copy text (with the appropriate software). Scanners look and operate much like personal copiers and share much of the same technology. Just place a sheet of paper on the glass (or platen) and push a software button, and the scanner copies your image to the computer.
As with copiers, the outward simplicity is deceptive. Inside the scanner is a linear array called a charge coupled device (CCD), which is composed of about 2,600 photosensitive cells or elements. In most desktop scanners the CCD is only about an inch square and is fixed in place.
A light bar moves across the object being scanned (the target), and the light is reflected to the CCD by a system of mirrors. Each cell produces an electrical signal proportional to the strength of the reflected light that hits it. The signal which represents one pixel in the original images is converted into a binary number and sent to the computer. Darker portions of the target reflect less light and are given lower numbers.
Early scanning devices (1bit) scanners were limited to one value per pixel, either on or off, which represented black or white. Most desktop scanners today have CCDs that can differentiate 256 shades of gray or levels of brightness. That is about the same brightness range as a black and white photograph and roughly twice what the human eye can distinguish.
The resolution of a scanner is measured in dots (pixels) per inch (dpi).This is a fixed number based on the number of cells in the array and the total area scanned. For example, an array with 2,590 cells covering 8.5 inches gives a resolution of 300 dpi. The smaller the area, the better the resolution: Most color flatbed scanners today have resolutions of 300 or 400 dpi. Many scanners can produce lower resolutions with the proper software: Some offer increased resolution on one axis by taking readings more often as the light bar moves. For instance; a 300 dpi unit taking readings every 1/600 inch yields a 300 by 600 dpi image.
Interpolation is another method of improving resolution mathematically. An algorithm is used to calculate the value of a pixel placed between two that are actually scanned. If one of the pair were read as 8 and the other as 12, the algorithm would place the value between the pair as 10. The success of the final image depends on the type of material being scanned.
Since CCDs (charged coupled devices) capture brightness levels, it makes sense that they can be found in gray scale scanners, which are used for reproducing both continuous tone and line art originals. But how does a scanner capture color with the same CCD? The answer: by creating three separate versions of the image, one for each of the three primary colors of light reflected by the target. Most color scanners do this in three passes: One pass with a red filter records the red component; a second pass with a green filter records the green light; a final pass with a blue filter records the blue (see Figure1).
"Truecolor" scanners, record 8 bits of data for each pass, for a total of 24 bits of data per pixel. The result is a composite image with 16.8 million possible colors (24-bit color). There are other approaches. Color scanners like the Epson ES300C have three different colored lights, which flash in sequence as they travel down the platen (Figure 2). The third method for creating color scans is to use three CCDs, each collecting information about only one color. The HP ScanJet Ilc uses a twin lamp light bar, two sets of filters, and a three stripe CCD to allow single pass scanning (Figure 3).
There is no clear advantage to any one technique. None of the vendors can promise that the data gathered by their products for the three different colors will be aligned exactly (a process called registration). Beyond the problem of regular light-bar movement, many units have to start and stop as the amount of data generated exceeds the computer s ability to process it. If they are a pixel or two off, the results will be a little fuzzy. Significantly more mis-alignment would be unacceptable.
![]()