In the classification of high-dimensional hyperspectral images, only spectral information is not sufficient to obtain successful results when the number of training data is small. In this case, spatial information can be exploited as well as spectral information. For this purpose, we aimed to use spatial information obtained from the fuzzy C-means (FCM) algorithm and spectral information together with the help of composite kernels to classify hyperspectral images. The composite kernels obtained in experimental studies are used for classification purposes by using extreme learning machines (ELM) and support vector machines (SVM); in addition to that, the results were presented comparatively in the tables.