This paper introduces an image browsing system, based on simulated breeding methods, for "looking around" an image database interactively. First users are requested to input the significance of each image based on their own viewpoint; these values are then interpreted as the fitness of genes, and the parameters associated with image retrieval are interactively optimized by queue-based (pipeline-type) genetic algorithm, which this paper proposes as a fit model for simulated breeding. Based on the scenario and the model of image browsing, satellite cloud image database is used for the experiments and evaluation of image browsing, and the history of image browsing is analyzed by the order of similarity-based retrieval. Results obtained from experiments demonstrate the usability of the proposed system.

      @InProceedings{ kt:iec98,
      author = {Asanobu KITAMOTO and Mikio TAKAGI},
      title = {Interactive Image Browsing Using Pipeline-type Genetic Algorithm},
      booktitle = {Workshop on Interactive Evolutionary Computation},
      pages = {31-36},
      year = 1998,
      month = 3,
      doi = {},
      note = { (in Japanese)},
      }