We present an experiment of image annotating for photographs of one collection on base of ontology. A fragment of ontology, which consists of concepts of visual objects, their features and relations, is constructed using SemTalk2 software tool. Using this ontology there are prepared semantic annotations for photographs from the collection. Semantic annotations are ready for their further automatic processing. We consider them as input for automatic generation of photograph descriptions in a given NL. In this paper we discuss types of visual concepts, structure of Ontology and Image Models and their possible applications.