| One of the initial challenges in any Optical Music Recognition (OMR) system is
the treatment of the staves. For musicians, stavelines are required to facilitate
reading the notes. For the machine, however, it becomes an obstacle for making
the segmentation of the symbols very difficult. The task of separating
background from foreground figures is an unsolved problem in many machine
pattern recognition systems in general.
There are two approaches to this problem in OMR systems. One way is to try to
remove the stavelines without removing the parts of the music symbols that
are superimposed. The other method is to leave the stavelines untouched and
devise a method to segment the symbols (Bellini, Bruno & Nesi, 2001; Carter,
1989; Fujinaga, 1988; Itagaki, Isogai, Hashimoto & Ohteru, 1992; Modayur,
Ramesh, Haralick & Shapiro, 1993).
In the OMR system described here, which is part of a large document analysis
system, the former approach is taken; that is, the stavelines are carefully
removed, without removing too much from the music symbols. |
