Abstract: Lucky imaging is a high-resolution image restoration technology by selecting, registering and superposing a small number of lucky images from a large number of short-exposure images.It can effectively reduce the effect of atmospheric turbulence on image quality. However, the traditional CPU-based lucky imaging algorithm is difficult to achieve real-time. Taking advantage of the parallelism and flexibility of FPGA, this paper proposes a novel FPGA-based lucky imaging algorithm and built an FPGA experimental system. The algorithm adopts an image selection strategy only selecting a fixed number of lucky images and without sorting, and an image registration strategy based on the row and column coordinates of the selected images, which can effectively save algorithm processing time and hardware resources, and achieve the purpose of real-time lucky imaging. The new real-time lucky imaging algorithm can be implemented on small and medium-sized FPGA in a simple way.The high resolution image obtained by the real-time lucky imaging algorithmis the same as that processed by the traditional CPU-based algorithm. Experiments show that, for a lucky imaging processing of 2 000 frames of the input images of 128×128 pixels, the running speed of the FPGA-based algorithm is 27 times faster than that of the algorithm previously proposed by our laboratory, and over 150 times faster than that of the traditional lucking imaging algorithm based on CPU + MATLAB, the processing frame rate is up to 197 fps. The proposed algorithm and its FPGA implementation technology can be used to build a true real-time lucky imaging system.