In this paper, a two-level integral equation fast Fourier transform (IE-FFT) algorithm has been proposed to analyze full-wave electromagnetic scattering problems. In two-level IE-FFT algorithm, the problem domain is first partitioned into large cartesian cells. Next, the Green's function between two well-separated cells is approximated by interpolation technique. Different from Lagrange interpolation method used in the conventional IE-FFT algorithm, the radial basis functions (RBFs) are employed to reduce the number of interpolation points. To enhance the computational performance, each cartesian cell is further partitioned into smaller cells. The RBFs defined in each large cell are approximated by Lagrange function defined in each small cell. Numerical examples are presented to demonstrate the good accuracy and computational performance of proposed algorithm.