Optimal design of an off-axis four-mirror system based on computational imaging

The off-axis reflective system exhibits severe off-axis aberrations due to its non-rotational symmetry. Traditional design methods compensate for off-axis aberrations by complicating the mirror surfaces. To address this issue, this paper proposes optimizing the off-axis four-mirror system through computational imaging to achieve high-quality imaging. First, most of the current mainstream optical design software relies on the selection of the initial structure. We propose an improved particle swarm optimization algorithm (PSO-GA) combined with cosine annealing and a genetic algorithm (GA) to obtain the off-axis four-mirror initial structure for subsequent optimization. To improve the restoration effect and effectively compensate for off-axis aberrations, we introduce a generative adversarial network model (ES-GAN), which combines the attention mechanism and spatially gated feedforward network for image restoration. This method achieves high-quality restoration of blurred images without increasing the complexity of the surfaces. By comparing the peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) between the restored and original images, the results indicate that the PSNR of the restored image increased by 19.43%, and the SSIM increased by 19.32%. This verifies the effectiveness of the proposed computational imaging method and provides a reference for the simplification tasks of complex surface-type systems.