Abstract: Many methods of graphics recognition have been developed throughout the years for the recognition of pre-seg mented graphics symbols but very few techniques achieved the objective of symbo l spotting and recognition together in a generic case. To go one step forward through this objective, this paper presents an original solution for symbol spot ting using a graph representation of graphical documents. The proposed strategy has two main step. In the first step, a graph base representatio n of a document image is generated that include selection of description pri mitives (nodes of the graph) and organisation of these features (edges). In the second step the graph is used to spot interesting parts of the image that potenti ally correspond to symbol. The sub-graphs associated to selected zones are then su bmitted to a graph matching algorithm in order to take the final decision and t o recognize the class of the symbol. The experimental results obtained on differ ent types of documents demonstrates that the system can handle different t ypes of images without any modification.

Abstract: At the preceding GREC, we have proposed to use a "bag of symbols" formalism (similar to the bag of words approach) for the indexing of a graphical document image database. In this paper, we extend the proposed approach through the introduction of a rejection stage in the system. This rejection is based on the use of an original One Class Classifier. Some preliminary results are proposed.

Abstract: How to evaluate the existing recognition algorithms under a uniform platform is an interesting research topic. Most of the comparisons among the algorithms only take use of the dataset in one field and the recognition accuracy as the standard, it is difficult to make people select a suitable algorithm in practice. In this paper, we try to show the various aspects of the algorithms based on the performance when they are applied to the datasets from different application fields. At the same time, two new performance evaluation measures, homogeneity and separability, are proposed and implemented for the comparisons. Experimental results help the researchers understand the algorithms better.

Abstract: In this paper, a method for matching symbols in line-drawings is presented. Facing both segmentation and recognition of symbols is a difficult challenge. Starting from the results of a vectorization proce- dure, a visibility graph is built to enhance the main geometric constraints which were specified during the construction of the initial document. The cliques detection, which correspond to a perceptual grouping of primi- tives, is used in the system to detect regions of particular interest. Both opened and perceptually closed curves are identified from aggregation of cliques. Finally, the recognition stage uses an attributed edit distance technique to match approximated curves within the host attributed re- lation graph and the ones from a collection of symbols.