Hyper spectral Imaging produces an image where each pixel is having narrow spectral bands with plentiful spectral information. Spectral bands refer to the large number of measured wavelengths bands of Electromagnetic Spectrum. The large number of spectral bands in hyper spectral data increases the computational burden. So, dimensionality reduction through spectral feature selection thoroughly affects the accuracy of the classification. The applications of hyper spectral images require to process given data and achieve two fundamental goals: 1) detect and classify the constituent materials for each pixel in the scene; 2) reduce the data volume (dimensionality), without loss of useful information, so that it can be processed efficiently by a human. We used the technique of DRR (Dimensionality Reduction via Regression) an unsupervised method for dimensionality reduction.