Abstract: The 2-D frequency spectrum for the geosynchronous orbit synthetic aperture radar (GEO SAR) is complicated and the range and azimuth 2-D space-variant character cannot be handled by conventional imaging algorithm. Considering the imaging parameters are varying with the 2-D space-variant character, we propose a novel imaging algorithm based on a second order approximate model and a 2-D stolt interpolation in range frequency domain and azimuth time domain. By this method, the coupling between range and azimuth can be corrected. With regard to the large amount of calculation in interpolation, the Nonuniform Fast Fourier Transform (NUFFT) is adopted here to replace the traditional 2-D interpolation and inverse FFT (IFFT). Finally, the theoretical and simulation analyses show that the proposed algorithm can correct the 2-D space-variant in GEO SAR and quality measurement of the imaging result close to the theoretical values.