GLCM, Gabor, and morphology profiles fusion for hyperspectral image classification

A fusion method for combination of spectral and spatial features for classification improvement of hyperspectral images is proposed in this paper. Gray level co-occurance matrix (GLCM), Gabor filters, and morphology profiles are powerful tools for extraction of texture, shape, and size from the neighboring pixels. We study different combinations of theses spatial features with spectrum data and find the best choice for fusion of spectral and spatial features to increase the classification accuracy. Moreover, we assess the performance of PCA for feature reduction of fused feature vector in the best case. The experimental results on two real hyperspectral images show the good performance of proposed fusion method compared to other studied cases.