There are scores of commercial and shareware programs that read
many different symbologies, but searching around a few years ago,
the few open source
readers I was able to find didn't meet my needs. This project
aims to create a high performance, stable, robust library
component with supporting infrastructure that makes it easy to use
in a variety of applications. (and most importantly... "it
sounded like fun at the time" :)
How does it work?
A common design for a bar code "image scanner" is to apply
digital image processing techniques to an image containing a bar
code: exact details vary, but this usually involves several filter
steps to cleanup noise, sharpen and enhance contrast, edge
detection and shape analysis to determine symbol location and
orientation, etc. Finally the data is extracted from this
pristine image. All of these processing stages require CPU cycle
and memory resources and are often sensitive to various "filter
parameter" configurations which are difficult for end-users to
understand and setup.
The ZBar library uses an approach closer to that used by "wand"
and "laser" scanners: linear (1D) bar codes are designed to be
decoded by a simple light sensor passing over the light and dark
areas of a symbol. Taking advantage of this, the ZBar
implementation makes linear scan passes over an image, treating
each pixel as a sample from a single light sensor. The data is
scanned, decoded and assembled on the fly.
Taking a cue from modern processing paradigms, ZBar further
abstracts this idea into a layered streaming model. Processing is
separated into independent layers with well defined interfaces,
which can be used together or individually plugged into any other
system. A high-level description of the modules is provided
here:
video input
Abstraction of a video device which produces a stream of
images for scanning. The current release has interfaces to
video4linux (versions 1 and 2). Support is in progress for VfW
(Windows)
output window
Simple abstraction of a display output window that can
present a scanned image to the user and accept input in
response. To maximize flexibility, the window may be opened and
owned by the library, or attached to an application managed
window embedded in a GUI. The current release supports basic
X11 interfaces (XVideo and XImage). Support is in progress for
VfW (Windows).
image scanner
Makes scan passes over a two-dimensional image to produce a
linear stream of intensity samples. The input images may come
from the video input module, or any external image source (such
as an image file output by a flatbed scanner or digital camera).
This module also incorporates the optional inter-frame
consistency heuristic applied to a video stream.
linear scanner
Scans a stream of abstract intensity samples to produce a
"bar width" stream. The intensity samples could be pixel values
from the built-in image scanner, pixel values from an alternate
external image scanner, or even raw sensor samples from a
"decoder-less" wand or laser sensor. The bars are detected and
measured by applying some very basic 1D signal processing to the
input sample stream
decoder
The decoder searches a stream of bar widths for recognizable
patterns and produces a stream of completely decoded symbol
data. The current release implements decoding for EAN-13,
UPC-A, UPC-E, EAN-8, Code 128 and Code 39 symbologies. Support
is planned eventually for PDF-417 and EAN/UPC add-ons
processor
One potential drawback of a completely independent modular
approach is that it can take some coding to tie all of the
modules together, complicating even simple applications. The
high-level "processor" module connects all of the other modules
to flexibly support many common uses. For example, this makes
it easy to pop up a window (or not) and scan for bar codes from
video or image sources with very few lines of code. The
included sample applications: zbarcam and
zbarimg are two examples of how this can be done
(UTSL).
widgets
For applications which already have a GUI, it does not
always make sense to open a separate window for reading bar
codes. To facilitate tighter integration between the reader and
an existing GUI, the library also comes with ready-made
"widgets" for various popular toolkits (currently Qt4, GTK+-2.0
and PyGTK2). The test programs in the distribution are good
examples of how you can use these widgets to quickly incorporate
a bar code reader widget into your application.
language interfaces
With the performance sensitive image processing done in C,
library wrappers for Perl (currently available) and Python make
building a bar code application fast and easy!
plugin
Support has been started for a NPAPI (Mozilla) plugin, which
is supported by popular linux browsers and OpenOffice.org
(among others).
