Unsupervised multispectral image classification using MRF models and VQ method

An unsupervised contextual classification method using Markov random field (MRF) models and the vector quantization (VQ) is proposed. The VQ algorithm classifies the observed data preliminarily, and the contextual reclassification and the iterative classification follow. The proposed method was applied to a Landsat image to evaluate the classification accuracy.