Multi-sensor data classification in remote sensing using MRF regional growing algorithm

Abstract

This paper studies a multi-stage method using hierarchical clustering for unsupervised image classification to classify the land-cover using remotely-sensed data from multiple sensors. The multi-stage method performs region-growing segmentation using a hierarchical clustering procedure which makes use of the spatial contextual information by characterizing geophysical connectedness of digital image structure with Markov random field.