Study on the approaches for segmentation of hyper-spectral images based on energy-morphology and edge restriction

Concept and method of energy-morphological operations are simply expounded. The paper studies properties in hyper-spectral remote sensing data. The optimal bands are obtained by the correlation and information quantity, then gradient and edge are acquired by energy-morphological function, and then we can gain image constructed by gradients map overlaying edge map. Finally, by local minima values and watershed transform, complete images segmentation process. The researches and experiments indicate the approach combines a spectral part and a spatial part of hyper-spectral images; and provides foundation for further treatment of image.