The image restoration framework is illustrated in the diagram, which is composed of three main parts: the Monte Carlo simulation part, the image restoration part, and the parameter selection part. In the Monte Carlo simulation part, PSFs are generated under different observation conditions and sample images are convolved to generate blurred images as training data. The image restoration part restores these blurred images. Meanwhile, the parameter selection part evaluates the quality of the restored images and generates parameter dictionaries for the data generation part. To speed up the training process, we employ the MPI technology to run the data generation and image restoration parts in separate processors or computers.