1. Summary

This research proposes a new statistical image classification method based on the probaiboistic model of mixels or mixed pixels which appear most frequently on images with coarse spatial resolution such as remote sensing images. We are now developing this method under the name FCA(Fractional Component Analysis) in a broader framework.

2. Statistical Image Classification in the Presence of Mixels

The purpose of this work is to establish an image classification method that considers the spatial quantization effect of digital imagery, and its inevitable consequence, the presence of mixels. To achieve this goal, we propose new probabilistic models, namely the area proportion distribution and the mixel distribution. The former probabilistic model serves as the prior distribution of area proportions that reflects the internal structure of mixels, and is represented by Beta distribution. On the other hand, the latter probabilistic model is a totally unique model both in concept and its flat shape. Based on these probabilistic models, we claim that an image histogram should be modeled by the combination (mixture density model) of the mixel distributions for ``trough'' regions and the pure pixel distributions for peak regions. Then the expected area proportion is proposed for computing the area proportions of mixels. The experimental results on satellite image classification demonstrates that our probabilistic models work effective, both in terms of quantitative and qualitative evaluation, for images with the presence of mixels.

3. References (Complete List)

1. Asanobu KITAMOTO, "The Moments of the Mixel Distribution and Its Application to Statistical Image Classification", Advances in Pattern Recognition (SPR'00), Lecture Notes in Computer Science (LNCS) 1876, Amin, A. and Ferri, F.J. and Inesta, J.M., and Pudil, P. (Eds.), pp. 521-531, Springer, doi:10.1007/3-540-44522-6_54, 2000-8 [ Abstract ] [ Paper ]
2. Asanobu KITAMOTO, Mikio TAKAGI, "Area Proportion Distribution -- Relationship with the Internal Structure of Mixels and its Application to Image Classification", Systems and Computers in Japan, Vol. 31, No. 5, pp. 57-76, doi:10.1002/(SICI)1520-684X(200005)31:5<57::AID-SCJ6>3.0.CO;2-1, 2000-5 [ Abstract ]
3. Asanobu KITAMOTO, Mikio TAKAGI, "Image Classification Using Probabilistic Models that Reflect the Internal Structure of Mixels", Pattern Analysis and Applications, Vol. 2, No. 1, pp. 31-43, doi:10.1007/s100440050012, 1999-4 [ Abstract ]
4. Asanobu KITAMOTO, Mikio TAKAGI, "Image Classification Method Using Area Proportion Density that Reflects the Internal Structure of Mixels", Transactions of Institute of Electronics, Information, and Communication Engineers (IEICE), Vol. J81-D-II, No. 11, pp. 2582-2597, 1998-11 (in Japanese) [ Abstract ]
5. Asanobu KITAMOTO, Mikio TAKAGI, "Estimating the Area Proportions of Mixels Using Mixture Density Estimation with Mixel Densities", Transactions of Institute of Electronics, Information, and Communication Engineers (IEICE), Vol. J81-D-II, No. 6, pp. 1160-1172, 1998-6 (in Japanese) [ Abstract ]