Generalized gamma mixtures for supervised SAR image classification

In this paper we develop a new statistical model for supervised classification of high resolution synthetic aperture radar (SAR) amplitude images. This model is based on the recently proposed generalized gamma distribution (GΓD) for statistics of amplitude SAR images. In order to improve the fit of GΓD when dealing with inherently heterogenous high resolution SAR imagery, we model the statistics of thematic classes as mixtures of GΓD. This enables to consider not homogeneous thematic classes, which is an often requirement in practice. We complete the developed method by proving the identifiability of the developed GΓD finite mixture model and the consistency of the involved parameter estimation scheme (method of log-cumulants) for GΓD, which renders the developed approach mathematically correct. In order to improve the computational performance of the GΓD mixture estimation we suggest the use of an approximative solution of the equations involved, thus, avoiding time-consuming iterative processes. The accuracy of the developed approach is validated on a high resolution TerraSAR-X image and compared to related finite mixture-based SAR classification techniques.

[1]  Wen Hong,et al.  Generalized gamma distribution with MoLC estimation for statistical modeling of SAR images , 2007, 2007 1st Asian and Pacific Conference on Synthetic Aperture Radar.

[2]  S. Kotz,et al.  Statistical Size Distributions in Economics and Actuarial Sciences , 2003 .

[3]  E. Stacy.  A Generalization of the Gamma Distribution , 1962 .

[4]  Gabriele Moser,et al.  Enhanced Dictionary-Based SAR Amplitude Distribution Estimation and Its Validation With Very High-Resolution Data , 2011, IEEE Geoscience and Remote Sensing Letters.

[5]  Josiane Zerubia,et al.  Unsupervised parallel image classification using Markovian models , 1999, Pattern Recognit..

[6]  Kai-Sheng Song,et al.  Globally Convergent Algorithms for Estimating Generalized Gamma Distributions in Fast Signal and Image Processing , 2008, IEEE Transactions on Image Processing.

[7]  S. Quegan,et al.  Understanding Synthetic Aperture Radar Images , 1998 .

[8]  Gabriele Moser,et al.  High resolution SAR-image classification by Markov random fields and finite mixtures , 2010, Electronic Imaging.

[9]  G. Celeux,et al.  Stochastic versions of the em algorithm: an experimental study in the mixture case , 1996 .

[10]  Josiane Zerubia,et al.  Satellite image classification using a modified Metropolis dynamics , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[11]  J. M. Muñoz-Pichardo,et al.  A new condition for identifiability of finite mixture distributions , 2006 .

[12]  Nakagami Law,et al.  GAMMA MIXTURE MODELED WITH "SECOND KIND STATISTICS" : APPLICATION TO SAR IMAGE PROCESSING , 2002 .

[13]  Henri Maître,et al.  A new statistical model for Markovian classification of urban areas in high-resolution SAR images , 2004, IEEE Transactions on Geoscience and Remote Sensing.