Pyramidal contextual classification of SAR images

Pyramid-based classification techniques have shown great performance when compared to global classification methods. However, the accuracy becomes very low when spectral parameters only are considered. A new classification method is proposed which deals with integration of two algorithms: the hierarchical image segmentation by step-wise optimization to take into account the spatial context, and the Iterative Conditional Mode algorithm to classify the segmented image. After a comprehensive description of the new algorithm, its performance is analysed with SAR agricultural data.