A linear discrete ill-posed problem has a perturbed right-hand side vector and an ill-conditioned coefficient matrix. The solution to such a problem is very sensitive to perturbation. Replacement of the coefficient matrix by a nearby one that has less condition number is one of the well-known approaches for decreasing the sensitivity of the problem to perturbation. In this paper, we suggest some new regularization matrix to the Tikhonov regularization. These new ones are based on fractional derivatives such as Grunwald-Letnikov and Caputo and can cause to have more exact solutions.