Remote sensing has been providing solutions in a variety of sectors as a result of the recent growth in technology and available data. Land cover mapping is the most widely used application of remote sensing, yet it has always been a challenging task owing to various complexities. Consequently, constant studies have been conducted for the improvement of land cover classification accuracy using new datasets or algorithms in various cases. Because of free availability and high temporal coverage, Landsat data series are frequently chosen in studies of land cover mapping. In addition, various spectral indices (SIs) have been developed to separate a single feature very efficiently. Some of the derived SIs were stacked with an original multispectral image in some studies, in expectation of better classification results. In this study, we investigate whether the stacking of layers with different SIs derived from reflectance data could improve the land cover classification of Landsat OLI image in a hilly region of South Korea. A decision tree was used for the selection of SIs that aid classification. For that, five supervised classifiers, namely, Mahalanobis distance (Mahd), maximum likelihood (ML), minimum distance to means (MinD), parallelepiped (PP), and support vector machine (SVM), in three cases of a study area were applied with the same training and validation data to compare the accuracy of the results for original and two derived composites. Out of 45 land cover cases, 28 cases showed improvements by layer stacking indices. PP showed improvement in all cases but at the cost of unclassified pixels. MahD and SVM showed improvement in most cases with higher classification accuracy. ML was unable to classify the composite with all derived bands. In conclusion, layer stacking of the derived bands, even two normalized difference vegetation index and normalized difference water index, was able to improve the overall accuracy. Improving the accuracy of land cover maps would provide accurate information and is beneficial to authorities for a better understanding and analysis of the environment.