2004
DOI: 10.1007/978-3-540-28640-0_33
|View full text |Cite
|
Sign up to set email alerts
|

A Hybrid Approach to Detect Graphical Symbols in Documents

Abstract: Abstract. We propose to combine a feature descriptor method with a structural representation of symbols. An adaptation of the Radon transform, keeping main geometric transformations usually required for the recognition of symbols, is provided. In order to improve the recognition step we directly process on the grey level document. In this perspective, a three-dimensional signature integrates into a same formalism both the shape of the object and its photometric variations. More precisely the signature is compu… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
6
0

Year Published

2006
2006
2015
2015

Publication Types

Select...
4
2
1

Relationship

4
3

Authors

Journals

citations
Cited by 11 publications
(6 citation statements)
references
References 12 publications
0
6
0
Order By: Relevance
“…The work described in this paper can be pursued following several leads. First, concerning the symbol detection problem, the subgraph isomorphism approach could be applied to vector or contour-based structural descriptions of symbols and documents as the ones proposed by [37] or [41], in order to overcome some of the limits of region-based descriptions. At the same time, in order to manage possible segmentation errors which may result in merged or split vertices, operators establishing distances for one-to-many or Table 6: Precision and recall of the symbol spotting system implementing the rejection strategy based on the maximization of the F-measure many-to-one mappings [42] should be considered.…”
Section: Resultsmentioning
confidence: 99%
“…The work described in this paper can be pursued following several leads. First, concerning the symbol detection problem, the subgraph isomorphism approach could be applied to vector or contour-based structural descriptions of symbols and documents as the ones proposed by [37] or [41], in order to overcome some of the limits of region-based descriptions. At the same time, in order to manage possible segmentation errors which may result in merged or split vertices, operators establishing distances for one-to-many or Table 6: Precision and recall of the symbol spotting system implementing the rejection strategy based on the maximization of the F-measure many-to-one mappings [42] should be considered.…”
Section: Resultsmentioning
confidence: 99%
“…In other words, a graphical element can be defined as a symbol with a particular meaning in the context of a specific application domain (Lladós et al, 2002). Our goal is very similar to spotting, but we view this as a kind of retrieval (Delalandre et al, 2010;Qureshi et al, 2008;Tabbone et al, 2004) which is basically guided by user queries.…”
Section: Motivationmentioning
confidence: 99%
“…We have worked on signatures based on force histograms [38] and on the Radon transform [36,37], which enable us to localize and recognize complex symbols in line-drawings. We are currently working on extending the Radon signature to take into account photometric information, in order to improve the results when retrieving similar symbols in graphical documents [39]. By using a higher-dimensional signature, we are able to include both the shape of the object and its photometric variations into a common formalism.…”
Section: Symbol Spottingmentioning
confidence: 99%