Abstract: Atmospheric turbulence causes ground-based telescope imaging blur, the K-T high spatial image reconstruction algorithm is an effective way to solve this problem and the speckle imaging technique is acommon one. However, some factors such as large data and complex calculation, cause the reconstruction to be time-consuming. In this paper, we focus on the research of high spatial image reconstruction under the CPU and GPU hybrid architecture. By using MPI-CUDA hybrid parallel technology. we implement a high spatial resolution solar image reconstruction system based on speckle imaging techniques in stand-alone GPU environment. The experimental result shows that the processing speed of image sub-block has been improved significantly compared with the one only using MPI parallel, and the acceleration ratio of the whole process has reached 2 with 8 sub-processes. The experimental results show that the effectiveness of MPI-CUDA hybrid parallel can provide a good reference for large-scale computing tasks in astronomical research.