Spectral-spatial classification of high dimensional images using morphological filters and regression model

The integration of spectral and spatial features can significantly improve the hyperspectral image classification. In this paper, a spectral-spatial classification method is proposed which extracts the spatial features from the morphology profile of image using a regression model. It obtains the relationship between the neighbouring pixels in a spatial window of morphology profile obtained from the hyperspectral image using a regression model and considers the regression coefficients as extracted spatial features. Then, the morphology features, the regression coefficients and the original spectral features are fused together to form the final feature vector for classification. The experimental results show the superiority of proposed method compared to some other methods from the classification accuracy point of view.