DIPLI: Deep Image Prior Lucky Imaging for Blind Astronomical Image Restoration

2025-03-20

AI Summary

To address overfitting, artifact generation, and optimization instability in deep learning-based blind restoration of astronomical images—caused by scarce training data—this paper proposes an unsupervised multi-frame joint reconstruction framework. The method introduces a novel Markov stochastic optimization mechanism by coupling Langevin dynamics sampling with variational input regularization. It further enforces multi-frame consistency through synergistic integration of total-variation-based network (TVNet) priors and back-projection constraints. Crucially, the approach requires no ground-truth paired data or pretraining, effectively suppressing spurious noise learning and structural artifacts under low-data regimes. Experiments on diverse astronomical image datasets demonstrate that the method achieves a 2.1 dB PSNR improvement over the original Deep Image Prior (DIP), outperforming state-of-the-art diffusion models, vision Transformers, and classical Lucky Imaging. Moreover, training stability is enhanced by a factor of 3.8.

Technology Category

• Reasoning under Uncertainty: Stochastic Optimization
• Computer Vision: Learning & Optimization for CV
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Contemporary image restoration and super-resolution techniques effectively harness deep neural networks, markedly outperforming traditional methods. However, astrophotography presents unique challenges for deep learning due to limited training data. This work explores hybrid strategies, such as the Deep Image Prior (DIP) model, which facilitates blind training but is susceptible to overfitting, artifact generation, and instability when handling noisy images. We propose enhancements to the DIP model's baseline performance through several advanced techniques. First, we refine the model to process multiple frames concurrently, employing the Back Projection method and the TVNet model. Next, we adopt a Markov approach incorporating Monte Carlo estimation, Langevin dynamics, and a variational input technique to achieve unbiased estimates with minimal variance and counteract overfitting effectively. Collectively, these modifications reduce the likelihood of noise learning and mitigate loss function fluctuations during training, enhancing result stability. We validated our algorithm across multiple image sets of astronomical and celestial objects, achieving performance that not only mitigates limitations of Lucky Imaging, a classical computer vision technique that remains a standard in astronomical image reconstruction but surpasses the original DIP model, state of the art transformer- and diffusion-based models, underscoring the significance of our improvements.

Problem

• Enhance Deep Image Prior for blind astronomical image restoration.
• Address overfitting and noise in deep learning for astrophotography.
• Improve stability and performance in astronomical image reconstruction.

Innovation

• Enhanced DIP model with Back Projection and TVNet
• Markov approach with Monte Carlo and Langevin dynamics
• Improved stability and reduced overfitting in training