The development of a human-centered object-based image retrieval engine

The development of a new object-based image retrieval (OBIR) engine is discussed. Its goal was to yield intuitive results for users by using human-based techniques. The engine utilizes a unique and efficient set of 15 features: 11 color categories and 4 texture features, derived from the color correlogram. These features were calculated for the center object of the images, which was determined by agglomerative merging. Subsequently, OBIR was applied, using the color and texture features of the center objects on the images. The final OBIR engine, as well as all intermediate versions, were evaluated in a CBIR benchmark, consisting of the engine, the Corel image database, and an interface module. The texture features proved to be useful in combination with the 11 color categories. In general, the engine proved to be fast and yields intuitive