While collaborative tagging is attractive because of their easy and simple way to tag and search, tag based search suffers from no semantic descriptions for contents. Semantic web technologies provide a new way to annotating and retrieving images. We address the problems of both folksonomy and semantic web annotation systems for image retrieval in particular and argue that this can be improved by giving meanings to users' tags. This paper considers using both minimized domain-specific ontology and generic ontology. We show how less detailed ontologies can help end users to annotate and search with the simple user interface. Our prototype system has the capability to describe the semantics for contents of an image. Also, no previous extra work to create instances according to the ontology for later annotations is needed.