A Bayesian Learning Based Approach for Clustering of Satellite Images

This paper presents a technique for performing unsupervised clustering of satellite images using a unique ´sampling-resampling´ based Bayesian learning method. The multi-band pixel values of the satellite image are expected to form a certain number of clusters. The parameters of these clusters are learnt using a Bayesian approach. This technique is unsupervised in the sense that no separate training images are required to initialize the model parameters. Learning of cluster parameters and classification of pixels are done simulaneously. Parameter values obained using Bayesian techniques are expected to be more accurate, hence leading to better classification results, as compared to classical frequentist techniques. Also, the presented ´sampling-resampling´ based approach of performing Bayesian learning suggests computational simplicity and ease of implementation.