Doubly selective channels can cause severe perfor- mance degradation in orthogonal frequency division multiplexing (OFDM) systems, introducing inter-carrier interference (ICI) at the receiver. In such cases, equalization schemes which require matrix inversion are prohibitively complex for large OFDM symbol lengths. In this paper, we propose two low-complexity iterative successive interference cancellation schemes, applying Krylov subspace optimization methods. We first derive a reduced- rank preconditioned conjugate gradient (PCG) algorithm in order to estimate the equalization matrix with a reduced number of iterations. We then develop an improved PCG algorithm with the same complexity order, using the Galerkin projections theory. As verified via simulations, the proposed schemes may offer near optimal performance with reduced computational complexity.