Zusammenfassung: |
Due to their enormous sizes and their rapid growth, indexing modern image databases cannot be done manuallyanymore. Consequently, adaptive interactive image retrieval has become an important topic of database as well as of com-puter vision research. The standard approach to theme-related image retrieval is to iteratively browse the content: given aselection of images, the user rates their relevancy for his query. Using techniques from pattern recognition and computervision, this feedback is applied to narrow the choice of motifs presented in the next iteration. An important goal for thepractical application of such retrieval systems is to provide them with user interfaces that are intuitive to use and easy tolearn. In this contribution, we present a system that considers both these aspects. It applies adaptive retrieval techniquesand disposes of multimodal input devices which facilitate the interaction. A series of usability experiments carried outwith our system underlines that in combination these features allow for comfortable and successful interactive imageretrieval. |
