Study of Bayes Theorem for Classification of Synthetic Aperture Data

In this paper we describe a probabilistic approach of image classification based on Bayes classifier for SAR (Synthetic Aperture Radar) data. It classifies the images of different polarizations as well as the combination of these polarizations. The classification is carried out on the HH, HV and VV polarized image, their combination HH-HV, HH-VV and HV-VV and considering all the three polarized images, i.e., HH-HV-VV. It is an supervised classification approach which obtains the training data set from the region under observation. The results obtained are compared on the basis of better distinguishing ability of different classes by the classifier. The proposed classifier incorporates the use of Bayesian function for the evaluation of PDF (Probability Density Function) and with the consideration of PDF the decision function allocates a particular class to a pixel which better suits to that pixel.