Reconnaissance de symboles sans connaissance a priori

Abstract : A complete system able to find symbols in graphical document without a priori knowledge is proposed here. In a first place, this system is based on a structural method able to put in stress regions that may contain symbols. The document is represented by chain points that will be merged following a defined criteria. These merges allow potential symbols to be reconstructed. A descriptor is then calculated for each potential symbols, and the recognition can take place when the user submit a request. In order to speed up the retrieval, an indexing method based on BIRCH has been proposed by using a robust descriptor. Then we show that by combining filtering rules based on simple descriptors, we can rise the speed of the retrieval.
Document type :
Theses
Complete list of metadatas

Cited literature [54 references]  Display  Hide  Download

https://hal.univ-lorraine.fr/tel-01752768
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 1:57:20 PM
Last modification on : Tuesday, April 24, 2018 - 1:30:12 PM

File

2006_ZUWALA_D.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01752768, version 1

Collections

Citation

Daniel Zuwala. Reconnaissance de symboles sans connaissance a priori. Autre [cs.OH]. Institut National Polytechnique de Lorraine, 2006. Français. ⟨NNT : 2006INPL071N⟩. ⟨tel-01752768⟩

Share

Metrics

Record views

28

Files downloads

28