This paper considers interference mitigation with power allocation in cognitive radio systems. In such a system, cognitive radio is designed to transmit data to its corresponding users without causing any interference to the current primary transmission. To avoid these interference between these links, the secondary transmitters sense the primary user's unused eigenvector space and reuse these spatial hole resource in the same resource block. Moreover, an interference suppression scheme based on matrix distance is presented. Due to limit of the numerical times, the overleft leakage interference in the desired signal space is still exist. To improve the secondary users and total performance, we jointly optimize the precoding matrix, decoding matrix and power loading matrix under the constraints of transmission power of each transmitter. Since the problem is non-convex, the iterative technique and Cauchy-BuniakowskySchwarz theorem are used. Extensive computer simulation are provided and demonstrate that our proposed algorithm extremely outperforms the conventional scheme.