Extraction of knowledge from Tunisian historical mosaics using fuzzy logic and semantic concepts similarity measure to create a fuzzy metadata

The objective of this this research is to attain a semantic interpretation of images of historical mosaics. We use fuzzy logic techniques and a modified semantic similarity measure to extract knowledge from the images. The extracted knowledge provides experts and laypersons with an intuitive way to describe and to query the images in the database. The knowledge we extract from a mosaic image consists of (a) the fuzzy spatial relationships between objects in the mosaic image encoded using fuzzy inference and if-then rules complemented with some semantic rules, and (b) the semantic classification of mosaic using a modified semantic similarity measure. The mosaic classification can accelerate the retrieval process or can be used as a visualization method to the returned result. The system was tested using a database of 100 mosaics. We have compared spatial relationships extracted automatically with manual predefined ones. So, the automatic extraction of spatial relationships gives a success rate equal to 82%. Also, we have classified 76% of image into a correct semantic class, this rate can reach 87% if we take into consideration that the pertinent semantic class is one among two or more classes.