A new technique Multiple-feature-based adaptive sparse representation (MFASR) has beendemonstrated for Hyperspectral Images (HSI's) classification. This method involves mainly in foursteps at the various stages. The spectral and spatial information reflected from the originalHyperspectral Images with four various features. A shape adaptive (SA) spatial region is obtained ineach pixel region at the second step. The algorithm namely sparse representation has applied to get thecoefficients of sparse for each shape adaptive region in the form of matrix with multiple features. Foreach test pixel, the class label is determined with the help of obtained coefficients. The performances ofMFASR have much better classification results than other classifiers in the terms of quantitative andqualitative percentage of results. This MFASR will make benefit of strong correlations that areobtained from different extracted features and this make use of effective features and effective adaptivesparse representation. Thus, the very high classification performance was achieved through thisMFASR technique.