GSNR constructs a null-restricted graph Laplacian and projects onto its smoothest modes to regularize only the null-space part of inverse problem solutions, yielding up to 4.3 dB PSNR gains when plugged into PnP, DIP, and diffusion solvers.
Deep null space learning for inverse problems: convergence analysis and rates.Inverse Problems, 35(2): 025008
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GSNR: Graph Smooth Null-Space Representation for Inverse Problems
GSNR constructs a null-restricted graph Laplacian and projects onto its smoothest modes to regularize only the null-space part of inverse problem solutions, yielding up to 4.3 dB PSNR gains when plugged into PnP, DIP, and diffusion solvers.