Binarization of Matrix Codes - Hierarchical Neural Networks for Image Interpretation - page 155

8. Binarization of Matrix Codes
In this chapter, the binarization of matrix codes is investigated as an application
of supervised learning of image processing tasks using a recurrent version of the
Neural Abstraction Pyramid.
The desired network output is computed using an adaptive thresholding method
for images of high contrast. The network is trained to iteratively produce it even
when the contrast is lowered and typical noise is added to the input.
8.1 Introduction to Two-Dimensional Codes
Two-dimensional codes are an extension to one-dimensional barcodes, which have
been used for many years to mark items with machine readable numbers. In one-
dimensional codes the bits are represented by vertical black or white bars of variable
width. Figure 8.1(a) shows Code 39 [5], an example of a two-width code.
The first truly two-dimensional bar code was developed by David Allais at In-
termec Corporation in 1987 for space applications. As can be seen in Fig. 8.1(b),
Code 49 [6] is a stacked barcode with multiple rows. Parity bits, as well as check
characters at the end of a line and the end of the code ensure reliable decoding. At
most 49 characters can be stored in one symbol.
Figure 8.1(c) shows an example of the Data Matrix [4] code, developed in the
late 1980s by International Data Matrix (USA). Data Matrix is a two-dimensional
matrix symbology containing dark and light square data modules. It has a finder
pattern of two solid lines and two alternating dark and light lines on the perimeter
of the symbol. Larger codes contain additional finder patterns within the symbol.
(a) (b) (c)
Fig. 8.1. Different codes: (a) Code 39: one-dimensional two-width barcode; (b) Code 49:
stacked barcode; (c) Data Matrix: matrix code.