Unsupervised multispectral image classification

This paper presents a new approach to unsupervised classification for multispectral imagery. It first uses a Gaussian pyramid multi-resolution technique to reduce image size from which the pixel purity index (PPI) is implemented to find regions of interest (ROIs) with PPI counts greater than zero. These PPI-found samples are further used as support vectors for a support vector machine (SVM) to classify data. The resulting SVM-classified data samples are further processed by a new designed iterative Fisher´s linear discriminate analysis (IFLDA) which implements FLDA in an iterative manner to refine classification results. The experimental results show the proposed approach has great promise in unsupervised classification.