Following its growth from a plate solver into a comprehensive imaging engine, Seiza 0.18 refines its linear processing, stacking, and analysis capabilities. Under the hood, this release introduces on-disk stacking checkpoints, resilient calibration-master workflows, and a brand new star measurement subsystem that detects camera sensor tilt.

Resumable Live Stacking Checkpoints

The incremental LiveStacker engine (the core of seiza-stacking) has been updated to handle long-running, multi-night, or interrupted sessions gracefully.

A live-stacking session can now be atomically checkpointed to a versioned .seiza-stack context file on disk. This file preserves the full state of the stack—including registration reference points, per-channel normalization scales, and delta-sigma rejection history. You can close the application or restart the computer, reopen the context later, and resume stacking incoming frames without resetting the existing integration or losing the rejection math.

Resilient Calibration Workflows

Building calibration masters is historically fragile. A single corrupted flat frame or a dark frame with an mismatched exposure duration can easily ruin an entire master.

The new seiza-calibration workflow prioritizes consistency over rigid execution:

  • Isolate Disagreements: When processing a folder of bias, dark, dark-flat, or flat frames, Seiza evaluates each input against the collective statistics of the stack. If an individual frame disagrees or is corrupted, Seiza isolates it and continues building the master instead of throwing a hard error and aborting the entire run.
  • Audit Metadata: Skipped or problematic inputs are reported explicitly in the output metadata, and the resulting master FITS/XISF preserves the exact provenance of which files contributed to the integration.
  • Pedestal and Matching: leading up to this, the engine integrated frame matching, signal-to-noise ratio (SNR) analysis, and pedestal fitting to protect deep stacks from clipping shadow detail.

Star Measurement and Sensor Tilt Analysis

The legacy star detector has been replaced by seiza-stars, a dedicated Rust crate that handles star detection, sub-pixel centroiding, and PSF fitting.

The headline feature of the new crate is native sensor tilt analysis. By constructing a Delaunay triangulation of detected stars across the frame, Seiza analyzes variations in star shape, elongation, and FWHM across the focal plane. It calculates the exact orientation and magnitude of camera sensor tilt (how flat the sensor is relative to the telescope’s light cone) to help you isolate mechanical sag or tilt.

This analysis runs entirely locally and is fully exposed to:

  • Rust: Through the native seiza-stars and seiza-imgproc crates.
  • Python: Via the compiled binary wheels in seiza-py.
  • Native Apps: Over the unified seiza-cabi C-compatible library.

Desktop Apps Move to the 0.18 Core

The native Mac and Windows desktop applications both move to the v0.6 branch, linking the latest Seiza 0.18 engine over seiza-cabi:

  • Seiza for Mac (v0.6.2): This native Swift application fully integrates Finder thumbnail extensions and macOS Quick Look. You can browse FITS and XISF folders with instant, accurate file manager previews, and Quick Look fits the frame dynamically as you resize its window. The application includes reorderable stretch editors, radial-basis background gradient extraction, and local blind plate solving.
  • Seiza for Windows (v0.6.1): This native WinUI 3 application features GPU-accelerated panning and zooming, per-channel color rendering, deconvolution, and built-in catalog setup and repair. Release packages are fully signed with StackFoundry LLC’s Authenticode certificate.

Availability

Seiza 0.18 is available as Rust crates on crates.io, Python wheels on PyPI, and binary CLI executables on GitHub. Desktop applications are hosted in their respective repositories: seiza-mac and seiza-win.