Read errors of unreliable, one-dimensional bar codes often caused interruptions in industrial production, prompting companies like Toyota and its subsidiary Denso Wave to develop as early as 1994 a new code for acquiring stock data. The new matrix bar code was designed to store more information than the traditional bar code and to stay legible, even if the label was dirty, wrinkled, or partially destroyed.

The quick response code, or QR code, comprises a matrix of square dots instead of the usual lines. Measuring up to 177 by 177 dots, the QR code encodes up to 4,296 characters, compared with a bar code that encodes just 13.

Thanks to numerous free reader apps for smartphones, QR codes have gained in popularity in recent years. Posters, catalogs, magazines, business cards, and even television screens display the small squares, offering additional information or URLs for microsites.

Data Grabber

If you want to encode your address into a QR code for a business card, you could try one of the many services on the Internet. However, restrictions typically apply. For example, some services allow only QR codes in a certain size or in limited quantities. Also, you never know for sure, what the services do with your data. For example, some sites explicitly allow sharing and selling the data to third parties in their terms of use, more or less pre-programming an increase in junk mail in the future. Fortunately, several QR code tools can help you avoid these problems in Linux.

On My Mark!

The quickest way to create a QR code is with the Qrencode [1] command-line utility. Any major distribution can install Qrencode via the package manager. The following command then creates a QR code containing the text "Hello World!" (Figure 1):

Figure 1: Qrencode encoding the string "Hello World!" as a QR code and storing it in the qrcode.png file.

$ qrencode -o qrcode.png 'Hello World!'

The generated QR code ends up in the qrcode.png file. If the file already exists, it's overwritten without prompting. If you have at least version 3.3.0 (qrencode -V), Qrencode can generate an EPS graphic,

$ qrencode -t EPS -o qrcode.eps 'Hello World!'

or ASCII output (Figure 2). In the QR code image, the software creates a white border the width of one dot. This facilitates the process of deciphering the code for programs or the smartphone later on. If desirable, you can increase or decrease the edge with the -m parameter; in the following example the border width would be 10 code pixels:

Figure 2: Outputting QR codes in ASCII characters just for fun. Each # corresponds to a dot.

$ qrencode -m 10 -o qrcode.png 'Hello World!'

If you are saving the QR code in PNG format, the -s parameter specifies the height of a black QR code pixel. By default, Qrencode draws every black dot three by three pixels. The program creates a slightly smaller QR code with quite a wide margin with the following command:

$ qrencode -s 2 -m 10 -o qrcode.png 'Hello World!'

A white border appears around the QR code that is exactly 20 screen pixels wide (10 QR code dots in width, with each dot two pixels). If you are creating EPS images, the only parameter available is -m; -s moves the entire QR code out of the image in this case. To control the resolution of PNGs further, use the -d parameter to define the dpi.

Tolerance

Besides the specified data, the QR code contains additional error correction information. If a portion of the image is damaged, it allows you to reconstruct the missing or illegible data. The more additional information the QR code contains, the more heavily damaged it can be without becoming useless.

Increasing the error tolerance increases the size of the image because you need more black dots (Figure 3). The QR code standard thus uses four levels of error correction: H level allows you to read all data if 30 percent of the QR code is destroyed, Q level if 25 percent is unreadable, and M level if just 15 percent is unintelligible. At the lowest level, L, only 7 percent of the data can be faulty. In Qrencode, the -l parameter selects the error correction level. The possible values are pretty much what you would expect: L, M, Q, and H.

Figure 3: Both QR codes contain the text Hello World!; the one on the left uses the highest error correction level, H, and therefore is more robust – but also larger.

The level you choose depends on the proposed use. If you expect the QR code to be exposed to external influences (e.g., wind, weather, graffiti) or if you are planning to manipulate the image deliberately (see the "Intruders" box), you will want the highest level, H. On a business card, the lowest level, L, is sufficient. If in doubt, simply omit the -l parameter. The program then uses level L by default.