SyQon
000° N180° S270° W090° E

PARALLAX

Advanced Deconvolution & Nebula Deblur

Explore Architecture
PHYSICS-COMPLIANT RESTORATION

Deblur Stars. Recover Nebulae.

Traditional deconvolution amplifies noise, while generative AI models risk hallucinating false details. SyQon Parallax introduces a physics-compliant, gated, coordinate-aware residual regression network. With only 1.3 million parameters, it maps and corrects spatially-variant optical distortions (coma, astigmatism, structural flexure) on stars while simultaneously extracting hidden fine structures, dust lanes, and nebulosities with absolute mathematical fidelity.

Architectural Foundation

The StellarDirectNet Core.

4.5M+ Parameter Deep Residual Backbone

StellarDirectNet Pro utilizes a deep residual coordinate-aware framework with Squeeze-and-Excitation (SE) attention, C=128 base capacity, and multi-scale feature aggregation. Optimized for consumer hardware workstations, it converges up to 10× faster via OneCycleLR scheduling. Built on a non-generative residual mapping network.

High Bit-Depth Topology

Fully preserves 16-bit and 32-bit floating-point FITS formats, ensuring zero quantization collapse.

Zero Checkerboard Artifacts

Employs bilinear upsampling followed by convolutions to guarantee smooth, artifact-free stellar profile reconstructions.

Stellar & Nebular Optimization

By minimizing the second-order gradient moments:Lanisotropy = |Mxx - Myy| + 2·|Mxy|the network circularizes star profiles while the edge/gradient loss sharpens gas boundaries and nebulosities without inventing pixels.

Spatially-Variant Coordinate Mapping

Concatenates normalized coordinate channels (x, y coords ∈ [-1.0, 1.0]) directly to the input tensor. This enables the network to implicitly map aberrations that increase radially toward sensor corners.

Seeing-Calibrated Dataset Regression

Instead of generating images from scratch, the model regresses an additive residual map constrained by λ = 0.35. Trained on high-fidelity astronomical datasets with regression modeled using real public seeing data from the last 5 years, preserving background noise and faint gas details (> 41 dB PSNR).

Params4.5M+
Bckg PSNR> 41dB
Seeing Scope5 Years
Observe Empirical Validation

Interactive Aberration Fix.

Select a target below and drag the slider to verify the deconvolution. Parallax circularizes stellar profiles and restores fine nebular textures while keeping background noise completely unaltered.

After
Parallax: Nebula Restoration
Before
Raw Astrophotography Data

Nebula Target 01 correction: Extracts fine gas details and dust structures from dark nebulosities while suppressing background noise without generating artifacts.

Target: DEEP SPACE NEBULA FIELD 01
Calibration: 32-BIT FLOATING-POINT FITS DATA
Optical: RAW TELESCOPE SENSOR EXPOSURE

Acquire the logic.

Deploy SyQon Parallax in your local processing environment. Select your build below to gain immediate download access. Note: Parallax for PixInsight is fully compatible with both Apple Silicon and macOS Intel architectures.

Querying licensing server...

AI Deconvolution & Astronomical Stellar Restoration Hub

Parallax stands at the forefront of physics-compliant neural deconvolution for astronomical astrophotography raw data. Designed as a highly robust deep residual coordinate-aware network, it is built to circularize stellar profiles, correct optical aberrations (including comatic aberrations, astigmatism, and guiding flexure), and restore nebular gas lanes.

Whether you search for AI deconvolution, deconvoluzione AI, deconvolución con inteligencia artificial, ordéconvolution astronomique, Parallax provides a mathematical model tailored to sharpen cosmic details without introducing non-physical hallucinations.

Fully compatible as a plugin or processing module with popular software like PixInsight, Siril, and AstroPixelProcessor, our model preserves raw 16-bit and 32-bit FITS formats to maintain scientific data integrity.

Multilingual Index & Reference Guide / Glossario Deconvoluzione AI

Italiano (IT)Deconvoluzione AI per astrofotografia, deconvoluzione stellare, deconvoluzione immagini astronomiche, deconvoluzione PixInsight, deconvoluzione Siril, eliminazione sfocatura, recupero dettagli nebulosa.
Español (ES)Deconvolución de imágenes astronómicas con inteligencia artificial, deconvolución de estrellas mediante IA, restauración de nebulosas, corrección de coma y aberraciones ópticas en astrofotografía.
Français (FR)Déconvolution astronomique IA, déconvolution des étoiles, correction de la distorsion optique par réseaux de neurones, restauration de nébuleuses et réduction de flou.
Deutsch (DE)Astronomische KI-Dekonvolution, neuronale Bildschärfung für Astrofotografie, Beseitigung von Sternaberrationen, Dekonvolutions-Modell für Deep-Sky-Aufnahmen.
English (EN)Stellar deconvolution, astronomical AI deconvolution, neural deblurring model, cosmic image restoration, non-generative deconvolution software, PSF correction.
Other Terms & VariantiDeconv AI, deconvolution, deconvoluzione, deconvolucion, dekonvolution, deblurring, AI deblur model, neural deblurring, PSF restoration.