Abstract: Mathematical model of multi-carrier resource allocation in cognitive OFDM is constructed in Cognitive Radio Networks (CRN). Multi-carrier power allocation with the target of maximizing the transmit capacity of cognitive user is proposed under the condition of primary user’s interference constraint. Traditional water-filling algorithm is regarded as the optimal sub-carrier power allocation algorithm for single user. In this paper, two improved multicarrier power allocation algorithms based on water-filling are proposed. One of the improved algorithms can quickly determine the sub-carriers that do not require power injection by rough estimation of water level. The other one can directly determine the sub-carriers that do not require power injection by linear calculation. And there is no need for this algorithm to calculate the value of water level though iterative method. Meanwhile, all of them do not make any interference to primary user, which shows they are practical and feasible in CRN. Numerical results indicate that, the transmit capacity with the proposed power allocation algorithms are superior to that with traditional water-filling scheme for secondary user. However, the proposed sub-carrier power allocation algorithms could reduce the system computational complexity greatly and has some adaptive features.