Curvilinear spotlight SAR (CSSAR) has a high degree of freedom and can be used for 3-D imaging and video SAR (ViSAR) persistent imaging. The direct and effective processing of CSSAR data is an important part of CSSAR applications. However, CSSAR has higher requirements for motion compensation, especially when the motion measurement is not accurate enough. In this paper, an extended polar format algorithm (EPFA) is proposed based on the non-uniform fast Fourier transform for CSSAR. First, a theoretical derivation and analysis of the 2-D space-varying phase error in CSSAR are carried out. Then, a 2-D autofocus algorithm is proposed, which takes into account the spatial variability of the phase error. The efficiency of EPFA embedded in 2-D autofocus processing is significantly higher than that of back projection. In addition, based on the principle of small-angle approximation and spatial frequency domain sub-aperture technology, a novel ViSAR image generation scheme for CSSAR is proposed, which can significantly reduce the number of redundant calculations. The proposed algorithm is verified experimentally using data with a bandwidth of 2.4 GHz.