Problem
Research questions and friction points this paper is trying to address.
image restoration
training-data-free
mixed degradation
implicit neural representation
sparse regularization
MG-SpaIR: Multi-grade Sparse-guided Implicit Representation for Training-Data-Free Image Restoration
🤖 AI Summary
This work proposes a training-free blind image restoration method for images degraded by a combination of blur, downsampling, noise, and missing pixels. Leveraging implicit neural representations (INRs), the approach introduces a coarse-to-fine multi-scale residual optimization framework augmented with an explicit ℓ₀-type sparse proximal regularization in the high-resolution image domain. This regularization effectively mitigates the spectral bias and artifacts inherent to INRs. By jointly modeling the mixed degradation process and structural image priors, the method achieves consistently superior performance over existing zero-shot techniques such as Deep Image Prior across multiple benchmarks, delivering stable, interpretable, and high-fidelity reconstructions with well-preserved fine details.
📝 Abstract
MG-SpaIR is a training-data-free framework for restoring a clean image from a single observation corrupted by a mixture of blur, downsampling, noise, and missing pixels. Building on implicit neural representations (INRs), we introduce a multi-grade coarse-to-fine residual hierarchy that progressively refines the reconstruction across resolution grades, improving representational fidelity and mitigating spectral limitations. To stabilize reconstruction optimization and suppress INR-induced artifacts, we further propose an explicit sparse proximal regularization (e.g., $\ell_0$-type) applied directly in the high-resolution image domain, which discourages spurious high-frequency patterns while preserving sharp structures. The resulting optimization is solved efficiently via a multi-grade proximal alternating scheme, and we establish convergence guarantees for the associated updates under standard regularity conditions. Experiments on mixed-degradation benchmarks demonstrate that MG-SpaIR consistently outperforms strong training-data-free baselines such as Deep Image Prior, providing a stable, interpretable, and data-efficient alternative to conventional learning-based restoration methods.
Innovation
Methods, ideas, or system contributions that make the work stand out.
implicit neural representation
training-data-free
multi-grade refinement
sparse proximal regularization
image restoration
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
