This paper focuses on a "mixel," which contains multiple classification classes inside a single pixel. To analyze the internal structure of mixels, we focus on the distribution of "area proportion," which represents the area proportion of each class within a mixel. This distribution has been (sometimes implicitly) assumed to be uniform distribution; however, this paper suggests that "Beta distribution" is a better model from the viewpoint that it can describe many types of distributions which are theoretically or empirically derived from simple figure models and fractal synthetic imagery. The proposed probabilistic model is integrated into a statistical image classification method proposed by authors, and the model using Beta distribution demonstrates good fitness to the image histogram, which results are then evaluated by information criterion.

      @Article{ kt:ieicetrans98-b,
      author = {Asanobu KITAMOTO and Mikio TAKAGI},
      title = {Image Classification Method Using Area Proportion Density that Reflects the Internal Structure of Mixels},
      journal = {Transactions of Institute of Electronics, Information, and Communication Engineers (IEICE)},
      volume = {J81-D-II},
      number = {11},
      pages = {2582-2597},
      year = 1998,
      month = 11,
      doi = {},
      note = { (in Japanese)},
      }