In digital protection systems, discrete Fourier transform (DFT) technique is the most widely used frequency analysis tool for extraction of fundamental and harmonic components of fault currents. However, some fault currents are accompanied by decaying direct current (DC) offset which affects the accuracy of the DFT-based methods as the polluted fault current with decaying DC offset is not a periodic signal. This matter may cause mal-operation of some protection schemes such as overcurrent and distance relays. In this study, to address this issue and decrease the estimation error, a novel technique is proposed to eliminate the effects of DC offset on DFT-based methods. In the proposed method, the decaying DC offset is estimated using one-cycle data windows of the fault current signal. By subtracting the estimated decaying DC offset from the fault current, the remaining signal becomes periodic. Thereafter, the DFT method can be effectively used for extraction of fundamental and other harmonic components. The performance of the proposed method is evaluated comprehensively by investigating different signals even in noisy conditions. Also, a prototype is conducted to validate the applicability of the proposed method in the real power system by real-time simulation.