From the segmentation to the complete reading of a mathematical document

• Autorzy: Toumit, J. Y. , Emptoz, H.
• Konferencja: Conference on Computer Graphics and Image Processing, GKPO' 98 (5 ; 18-22.05.1998 ; Borki, Poland)
• Języki publikacji: EN

Abstrakty

EN

Automatic reading of scolar manuals is an important problem for the editors; we are presently working in this context on the conversion of mathematical manuals into electronic documents. No research on the entire mathematical document seem to have been achieved until now, there has been only stidies on the formulas themselves. We therefore present the problem of reading such documents. Those documents contain two types of information of different natures: the next and the mathematical objects. To perform a better treatment on the text itself, we are leaded to separate those two types of information; in this article, we pay a paricular attention to this treatment which can be considered as a multi-language segmentation problem. Classical methods do not provide satisfactory results and we needed to introduce a new segmentation approach; it fills the document's surface using "propabation" methods around particularity specyfic points of the text or of the mathematical objects. We also analysed the constraints relative to the documents we have to deal with; in this context, we need to use the gray-level image without binarizing it. A method for segmenting words and characters using this gray-level image is presented and we then introduce tetrarization which leads to more reliable images than binarized images.

Słowa kluczowe

EN

automatic reading; mathematical formulas; multilingual segmentation; grayscale image

• Czasopismo: Machine Graphics and Vision
• Rocznik: 1998
• Tom: Vol. 7, No. 1/2
• Strony: 485-504
• Opis fizyczny: Bibliogr. 22 poz.

Twórcy

autor

Toumit, J. Y.,

autor

Emptoz, H.,

• Laboratoire de Reconnaissance de Formes et Vision Bat 403 INSA de LYON 20, av. A. Einstein 69621 Villeurbanne Cedex