Unsupervised classification of KOMPSAT EOC imagery based on independent component analysis

Independent component analysis (ICA) is a recently well known stochastic method for learning robust image filters, which transform the texture to meaningful features. Based on high order statistics, ICA learns both ICA filters and independent components at the same time. The proposed method firstly generates ICA filter by adopting the fast ICA algorithm from the KOMPSAT full scene and secondly classifies projected texture features under space. As there exist some finless texture areas on the gray KOMPSAT image, we organize new feature vector composed of large feature values and a normalized mean value of pixels in the window. A k-means clustering algorithm is used to classifying those new feature vectors. The proposed classification method shows sufficient performance for the KOMPST panchromatic image with 6.6 m ground sampling distance (GSD).