Hyperspectral images are the treasure of information since it contains hundreds of spectral bands. Classification of Hyperspectral images is the process of identifying the components in each pixel. For this purpose, the pure and mixed pixels of the image should be identified and the endmember signatures and components are identified with the help of spectral libraries. In this paper, it is attempted to identify the minerals in the hyperspectral data ‘Cuprite’, that covers the Cuprite mines in Las Vegas, Nevada, the U.S. The pure pixels are identified by using PPI algorithm and their endmember signatures are obtained. The abundance maps of mixed pixels are obtained by using LMM. The total variation based regularization and joint sparsity of abundance maps are exploited in this paper.