Satellite image classification is based on description, texture, or similarity of items or things. Satellite Image classification is a challenging task for machines. Satellite image classification is possible using characteristics, training sample, an assumption of the parameter on data, the pixel, the number of outputs for each spatial elements, spatial information, and multiple classifier approach. These approaches are summarized in this paper but the main objective of this paper to explore classification based on training sample, classification based on the training sample considers two approaches: supervised image classification and unsupervised classification.