Comparison of Two Goal-Oriented Methods for the Evaluation of the Text-Line Segmentation Algorithms

Text line segmentation process represents the key step in the optical character recognition. Hence, the efficiency evaluation procedure for text line segmentation algorithms is the challenge. Text line segmentation process is established by the algorithms application to the text dataset. Furthermore, two goal-oriented methods for the evaluation of the text line segmentation results based on extended errors type and binary classification are explained. The paper presents the main points of the provided analyses and results discussion. The results confirm the superiority of the extended errors type over binary classification evaluation method. Streszczenie. Przedstawiono analizę algorytmów segmentacji linii tekstu. Analizowano dwie metody – analizy błędu i binarnej klasyfikacji. Wykazano przewagę pierwszej z tych metod. (Porównanie dwóch algorytmów stosowanych do segmentacji linii tekstu)