Morphological filters for OCR: a performance comparison

Laurent Mennillo (1) , Jean Cousty (2) , Laurent Najman (2)

1 IP - Institut Pascal
2 LIGM - Laboratoire d'Informatique Gaspard-Monge

In this article is compared the ability of several morphological operators to improve OCR performance when used as preprocessing filters. An experiment on binary and greyscale images using the Tesseract OCR engine and morphological filters acting in complex, graph and vertex spaces has thus been performed and results in a good overall performance of complex and area filters. MSE measures have also been performed to evaluate the denoising ability of these filters, which again shows the good performance of both complex and area filters on this aspect.

Domaines: Traitement des images [eess.IV]; Traitement du texte et du document

Laurent Mennillo, Jean Cousty, Laurent Najman. Morphological filters for OCR: a performance comparison. [Research Report] LIGM - Laboratoire d'Informatique Gaspard-Monge. 2012. ⟨hal-00762631⟩