01
Observation first
The input observation—not an idealised target—defines the optics, sampling, noise and uncertainty that the method must handle.
SyQon Research / computational imaging
Methods for measurable astronomical images.
We study restoration and separation systems against quantities that matter in astronomy: point-spread function, flux, centroid, background statistics and uncertainty.
- Programmes
- 2 indexed
- Execution
- Local
- Primary domain
- Astronomical imaging
- Collaboration
- Open to institutions
Research index
Current programmes, their validation scope and publication state.
R-02
PLXNet + AADR
AADR: Adaptive astronomical restoration
A retained PLXNet backbone, multiscale corrector and three assessors that estimate candidate gains and can accept, stop or abstain.
- Field
- Deconvolution
- Evaluated with
- pixel · detail · stars · RGB bias
- Status
- Adaptive restoration
R-01
2026
Axiom V3: gated convolutions and multi-scale objectives for stellar separation
Architecture and training constraints for separating stellar profiles from extended deep-sky signal.
- Field
- Signal separation
- Evaluated with
- stellar flux · profile · background
- Status
- Published
Working principles
What we require from a research claim.
02
Measure more than appearance
Where possible we evaluate point-spread function, flux, centroid, residual structure and background statistics alongside visual inspection.
03
Separate research from product claims
Experimental configurations, synthetic tests and export-equivalence checks are identified as such. They are not presented as production benchmarks.
04
Run locally
Research becomes useful only when it can operate on real full-resolution data within practical memory and compute limits.
Validation language
Functional verification
Confirms graph export, shapes, finite values, tiling and numerical agreement. It does not measure restoration quality.
Synthetic validation
Uses known corruptions and targets to isolate behaviour. Its scope is limited by the simulator.
Observational validation
Tests full astronomical acquisitions, including cases not represented exactly by the training pipeline.
Institutional collaboration
Observatories and research groups can request a less restrictive research licence at no cost.
Tell us about the instrument, dataset and intended study. We will discuss access, reproducibility requirements and suitable evaluation criteria directly.