Geometric-Unconstrained Progressive Endmember Finding: Orthogonal Projection Analysis

  • Chein-I Chang


Chapter  8 develops a Causal Iterative Pixel Purity Index (C-IPPI) which can implement IPPI in a causal manner in the sense that each data sample vector is fully processed sample-by-sample one after another for a given fixed set of skewers to produce its own final PPI count. This chapter presents a rather different version of IPPI, referred to as Progressive IPPI (P-IPPI) by interchanging two iterative loops carried out in C-IPPI. In other words, P-IPPI can implement IPPI progressively in the sense that each data sample vector is processed by IPPI with growing sets of skewers skewer by skewer one after another. That is, for the number of skewers, K, each data sample vector will be processed by IPPI progressively as the number of skewers grows to from 1 to K and its PPI count will also be updated K times as every new skewer is processed. As a result, P-IPPI is quite different from C-IPPI from an algorithm design perspective. In particular, comparing C-IPPI, which requires the prior knowledge about the value of K and processes each data sample vector for all skewers sample by sample, P-IPPI processes all data sample vectors skewer by skewer with the value of K increased by one after each iteration is completed. Consequently, in theory P-IPPI can be implemented skewer by skewer progressively by increasing the value of K indefinitely. Nevertheless, for a fixed number of skewers, K, both C-IPPI and P-IPPI eventually produce identical results. In analogy with C-IPPI, the same major advantages that can be gained from C-IPPI are applied to P-IPPI. The only advantage provided by P-IPPI that C-IPPI does not have is that there is no need for P-IPPI to determine the value of K, while C-IPPI requires prior knowledge about the value of K.


Outer Loop Background Pixel Progressive Process Minimum Projection Panel Signature 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. Boardman, J.W. 1994. Geometric mixture analysis of imaging spectrometry data. International Geoscience Remote Sensing Symposium 4: 2369–2371.Google Scholar
  2. Chang, C.-I 2003a. Hyperspectral imaging: Techniques for spectral detection and classification. New York: Kluwer Academic/Plenum Publishers.Google Scholar
  3. Chang, C.-I 2003b. How to effectively utilize information to design hyperspectral target detection and classification algorithms, Workshop in honor of Professor David Landgrebe on Advances in techniques for analysis of remotely sensed data, NASA Goddard Visitor Center, Washington DC, October 27–28, 2003.Google Scholar
  4. Chang, C.-I, and Q. Du. 2004. Estimation of number of spectrally distinct signal sources in hyperspectral imagery. IEEE Transactions on Geoscience and Remote Sensing, vol. 42, no. 3, pp. 608–619, March 2004.Google Scholar
  5. Chang, C.-I, and A. Plaza. 2006. Fast iterative algorithm for implementation of pixel purity index. IEEE Geoscience and Remote Sensing Letters 3(1): 63–67.CrossRefGoogle Scholar
  6. Chang, C.-I, S. Chakravarty, and C.-S. Lo. 2010. Spectral feature probabilistic coding for hyperspectral signatures. IEEE Sensors Journal 10(3): 395–409.CrossRefGoogle Scholar
  7. Chen, H.M., C. Lin, S.Y. Chen, C.H. Wen, C.C.C. Chen, Y.C. Ouyang and C.-I Chang. 2013. PPI-iterative FLDA approach to unsupervised multispectral image classification, IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing 6(4): 1834–1842.Google Scholar
  8. Harsanyi, J.C., W. Farrand and C.-I Chang. 1994a. Detection of subpixel spectral signatures in hyperspectral image sequences. Annual Meeting, Proceedings of American Society of Photogrammetry and Remote Sensing, Reno, pp. 236–247.Google Scholar
  9. Harsanyi, J.C., W. Farrand, J. Hejl and C.-I Chang. 1994b. Automatic identification of spectral endmembers in hyperspectral image sequences. In International symposium on spectral sensing research ‘94 (ISSSR), San Diego, July 10–15, pp. 267–277.Google Scholar
  10. Winter, M.E. 1999a. Fast autonomous spectral endmember determination in hyperspectral data. In Proceedings of 13th international conference on applied geologic remote sensing, Vancouver, B.C., Canada, vol. II, pp. 337–344.Google Scholar
  11. Winter, M.E. 1999b. N-finder: an algorithm for fast autonomous spectral endmember determination in hyperspectral data. Image Spectrometry V, Proc. SPIE 3753: 266–277.CrossRefGoogle Scholar
  12. Wu, C.-C., H.M. Chen, and C.-I Chang. 2010. Real-time N-finder processing algorithms. Journal of Real-Time Image Processing 7(2): 105–129. doi: 10.1007/s11554-010-0151-z.CrossRefGoogle Scholar
  13. Xiong, W., C.-C. Wu, C.-I Chang, K. Kapalkis, and H.M. Chen. 2011. Fast algorithms to implement N-FINDR for hyperspectral endmember extraction. IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing 4(3): 545–564.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media, LLC 2016

Authors and Affiliations

  1. 1.BaltimoreUSA

Personalised recommendations