Hyperspectral data face challenges to image analysis, because of redundancy in the information, high data volume dimensionality problems and need for calibration. In this paper discusses various pre-processing method for hyperspectral image analysis. Removal of atmospheric effect before image analysis. Minimum Noise fraction is Transformation (MNF) technique is used to reduce the dimensionality of hyperspectral data. Endmember extractions is used to select endmember spectra for used in classification and advanced spectral analysis. Feature reduction is required to identify small set of bands for further analysis.