A classification enhancement in hyperspectral imagery using superresolution technique

In this paper an improved supervised classification technique through applying a superresolution method on remotely sensed hyperspectral images is introduced. Superresolution methods provide high-resolution images from a sequence of low-resolution frames. In the proposed technique, classification of the hyperspectral image is carried out using spectrally homogenous training classes of pixels. Low spatial resolution frames of different wavelengths of the hyperspectral image are fed to a quadratic programming based classification algorithm to enhance the spatial resolution of the classification process. The results show a better classification and an edge improvement. Target recognition is the main field which can benefit from this technique.