Construction of handwriting databases using transcript-based mapping
Bin Zhang, C. Tomai, S. Srihari, V. Govindaraju · 2004
A recognition-based system was developed for constructing handwriting databases. The system automatically recognizes the word and the character images in handwritten document images by applying a transcript mapping algorithm. The transcript-mapping process is modeled as an optimization problem involving multiple word-segmentation hypotheses, word recognition and word alignment. The extensive experiments show that a large number of character and word images can be automatically extracted with very high reliability.