Line Extraction from Unconstraint Handwritten Document Pages using Piece-wise Water-flow Technique.
Ram Sarkar, Subhadip Basu, Nibaran Das, Ayatullah Faruk Mollah, Mahantapas Kundu, Mita Nasipuri · Indian International Conference on Artificial Intelligence · 2009
To extract the lines from multi-skewed document images of handwritten Bangla or Roman text, a novel technique is described here. In this paper, at first, the document image is partitioned into number of vertical fragments. Then a hypothetical water-flow technique is applied separately in each vertically fragmented portion of the document images, from both left and right sides of the said partition. After identifying the touching segments, separation of the same is done and finally the water-flow technique is applied on the overall document image. The stripes of areas left unwetted on the image frame are finally labelled for extraction of text lines. To test the developed technique, Bangla and Roman scripts are used and the performance of the technique, as observed experimentally, is 95.54% for the document images of the said scripts.