Adaptive image refinement learns to allocate detail across scales
Learned Adaptive Multiresolution Diffusion Imaging
Machine Learning
Summary
Images can be made more efficient by only adding fine detail where it's really needed. The authors introduce a new method that uses machine learning to decide when and where to add those details, improving accuracy over traditional fixed rules. Their approach remembers the image structure, works at multiple scales, and performs better on tests without needing to be retrained for larger images. This helps create clearer images while using fewer resources.
What this means in practice
- •For medical imaging teams: Improve MRI image quality by adaptively assigning resolution based on learned policies to reduce error while controlling data size.
- •For satellite image processors: Enhance detail in high-resolution satellite images with efficient multiresolution techniques that transfer learning across different image sizes.
Authors
Christian Tantardini, Stig Rune Jensen, Roberto Di Remigio Eikås, Joakim Henrik Beck
Abstract
Adaptive multiresolution methods reduce representation cost by concentrating fine-scale degrees of freedom where needed, but their tree updates are usually governed by fixed local criteria. We introduce Learned Adaptive Multiresolution Diffusion Imaging (Learned AMDI), which preserves the AMDI fixed-tree propagator and hierarchy constraints while replacing the post-propagation selector with a shared local policy trained by proximal policy optimization. Regression tests reproduce deterministic AMDI trajectories to machine precision when identical trees are used. In the Haar implementation studied here, the deterministic one-step selector accepts no refinements in 54 decisions. Across nine held-out cases, Learned AMDI executes 393 refinements and reduces the mean terminal reference discrepancy from $0.17496$ to $0.13657$, while occupancy rises from $0.13737$ to $0.26660$. Step-resolved diagnostics reveal occasional small adaptation-energy increases; fixed-tree energy stability therefore does not guarantee monotonicity of the learned outer iteration. At comparable occupancy, a validation-tuned observed-detail threshold reaches a discrepancy of $0.13792$ with slightly better RMSE and SSIM, placing both methods on essentially the same accuracy--occupancy tradeoff. A decision-1-only control reaches $0.13742$, indicating that most of the improvement on this static benchmark arises from the initial allocation. The shared actor transfers without retraining to $64\times64$ and $128\times128$ images, improving reference discrepancy, RMSE, and SSIM relative to deterministic AMDI, while the frozen threshold rule remains competitive. Learned AMDI thus provides a hierarchy-constrained, resolution-transferable mechanism for adaptive allocation and clarifies the contribution of sequential decisions.