Hybrid CNN-Total Variation Deconvolution for Mitigating Atmospheric Seeing Effects in Ground-Based Astronomical Imaging

Ground-based telescopes often suffer from atmospheric seeing effects, which degrade the quality of astronomical images that are acquired. In this paper, we present an innovative hybrid model of Total Variation Deconvolution (TVD) and Convolutional Neural Network (CNN) architecture to tackle this challenge. The developed model was designed in the order of TVD as the first layer, three other layers of CNN followed, which created a four-layer structure. This is a strength-based model of both TVD and CNN, with the objective of image deblurring, clearness, sharpness, and fine detail preservation in the image. Performance measures of our hybrid model were evaluated using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) metrics along with visual inspection. The results show that the hybrid approach will significantly improve image quality over traditional methods, presenting a promising solution in improving resolution and detail in ground-based astronomical observations.