Superpixel based classification of hyperspectral images

Hyperspectral imaging captures a high number of spectrally narrow bands and provides advantages for image analysis applications such as identification and classification in particular. Hyperspectral images contain a large amount of bands. Processing these images causes the operation load substantially. Improved methods for the classification of hyperspectral image, can not succeed due to the multidimensionality. To overcome this disadvantage made size reduction and to reduce the number of bands. In this study, to hyperspectral image to be consistent with the human visual system, band gaps are selected which red (R), green (G) and blue (B) corresponding to the wave length. In this paper, superpixel approach is proposed to improve the classification performance.