In this article, a novel imaging algorithm by combining two-step scaling transform (TSST) with structure-aided 2-D autofocus is proposed for the squint spotlight synthetic aperture radar (SAR). First, on the basis of planar wavefront assumption, a modified range-frequency linear scaling transform (MRFLST) and an azimuth-time nonlinear scaling transform (ATNST) are proposed to eliminate the coupling between range-frequency and azimuth-time of the received echo. Furthermore, to improve the efficiency, the MRFLST is implemented by using the principle of chirp scaling (PCS), which involves only complex multiplications and fast Fourier transforms (FFTs) without any interpolation, meanwhile, a constant scaling factor (CSF) selecting criteria is defined to avoid range spectrum aliasing. Then, to correct the phase error caused by the range measurement error and atmospheric propagation effects, the prior 2-D phase error structure implied in the TSST is analyzed. Finally, by integrating the derived 2-D phase error structure and range frequency fragmentation technique, a new 2-D autofocus algorithm is presented to improve the image quality. Simulated and real data experiments are carried out to verify the proposed algorithm.