This work presents a spectral image fusion approach from compressive projections based on the linear mixture model that exploits the endmember matrix low dimensional structure. The formulated inverse problem includes a total variation term over the abundance matrix to promote smoothness, but also a low rank term over the endmember matrix to promote the low rank structure. The optimization problem is solved using an alternating direction method of multipliers (ADMM) approach to independently estimate the abundance and endmember matrices. Simulations show that the fusion problem can be effectively solved from compressive projections, and the inclusion of the low rank regularization increases the reconstruction quality.