Abstract : In the quest for models that could help to represent the meaning of images, some approaches have used contextual knowledge by building semantic hierarchies. Others have resorted to the integration of images analysis improvement knowledge and images interpretation using ontologies. The images are often annotated with a set of keywords (or ontologies), whose relevance remains highly subjective and related to only one interpretation (one annotator). However, an image can get many associated semantics because annotators can interpret it differently. The purpose of this paper is to propose a collaborative annotation system that brings out the meaning of images from the different interpretations of annotators. The different works carried out in this paper lead to a semantic model of an image, i.e. the different means that a picture may have. This method relies on the different tools of the Semantic Web, especially ontologies.