AUTOMATED K-WISHART CLUSTERING OF POLSAR IMAGES

This paper presents our automatic image segmenta-tion method for Polarimetric SAR data [1] with recent improvements by using matrix log-cumulant methods throughout. The clustering algorithm utilises the full polarimetric matrix information, incorporates texture by modelling with a non-Gaussian distribution, and auto-matically determines the appropriate number of classes with goodness-of-fit statistical tests. The modelling is based upon the well known product model, with a Gamma distributed texture parameter, leading to the K-Wishart model for the covariance matrix. The automatic clustering is achieved through a modified Expectation Maximisation algorithm, with an additional Goodness-of-fit test allowing splitting and merging of clusters. The resulting image segmentation depicts the statistically sig-nificant clusters within the image. Real world examples are shown to demonstrate the technique