Modified Self-organizing Maps for Line Extraction in Digitized Text Documents

. Different authors have developed modifications of the Kohonen Self-Organizing Maps to solve known combinatorial optimization problems. In this paper a modification of the Kohonen Map is proposed to solve the detection of white inter-text spaces in a digitized plain text documents. The idea relies on the fact that line extraction problem has several features which match easily with Kohonen networks, although an adaptation to the problem of the original learning rule has to be made at first. A test with different digitized text images is performed showing the ability to segment lines.