A new channel estimation method for orthogonal frequency division multiplexing (OFDM) system with large subcarriers and serious intercarrier interference (ICI) is proposed. The channel frequency-domain ( CFD ) matrix of each delay path is factorized to the product of a diagonal delay matrix and a circular ICI matrix in this model. To reduce the coefficient number, the circular ICI ma- trix is squeezed by using Hamming-window as the reshaping pulse in the transmitter. Meanwhile, the elements of the diagonal delay matrix are approximated with a discrete prolate spheroidal basis ex- pansion model (DPS-BEM). A least-square (LS) estimator is used to estimate the reduced channel coefficients. The proposed method is theoretically derived and simulated. The simulation results in- dicate that the model has good performance and is appropriate for various channel environments. The method also has low complexity and good spectral efficiency.