Two different algorithms have been developped to insure an efficient segmentation of the characters.
The first one, very simple but very fast, is systematically applied and is sufficent to perform the segmentation of the characters in most practical cases.
The second one has been especially conceived so as to separate characters one from each others when they are surrounding themselves. This second algorithm is required, and appears to be sufficent, in order to maintain the overall recognition performance as this situation occurs, which is often the case in practice, when there is no constrains of writing on the original document. As this second method is more complex and more expensive in terms of computation time than the first one, its use is controled by the classification step, which allows to limit its frequency, and, by doing so, to optimize the average segmentation time.
Back to the Technologies page.
Back to the Character Recognition Homepage.
send coments to firstname.lastname@example.org.