Summary: In traditional synthetic aperture radar (SAR) systems, the classical matched filter is extensively used to achieve the two-dimensional pulse compression of distance and azimuth dimension. The bandwidth of the system is often limited and fixed. In some practical applications, the output signal-to-noise ratio and target delay resolution have different requirements. The exponential filter with variable exponents is a better choice than matched filtering. In this paper, based on the exponential filter, a new two-dimensional exponential pulse compression transform is applied to the high-resolution SAR imaging algorithm. The simulation results show that the algorithm has a narrower main lobe width and lower side lobe height than the traditional SAR algorithm in both distance and azimuth dimension.