Project site: https://openpipp.github.io/opipp/
Grab the latest self-contained build from Releases.
Unzip anywhere — no .NET installation needed. It contains the GUI (Opipp.App.exe) and the
command-line engine (Opipp.Cli.exe).
OPIPP (pronounced "oh-pip", Open PIPP) is an independent, fully managed continuation of
PIPP (Planetary Imaging PreProcessor) by Chris Garry. The name keeps the original's initials so
the lineage stays obvious, while everything in this repository — assemblies, namespaces, executables,
projects — is Opipp.*.
OPIPP is at 0.3.0-alpha. The engine is functionally complete and parity-tested against the original across 27 compared cases, with two declared differences, but one algorithm (surface-feature tracking) is still one pixel off on part of the real-data fixture, and the UI has not seen broad real-world use. Treat it as a working preview.
A new OPIPP extension, -derotate=on, corrects field rotation during final-image stacking — the
original PIPP aligns by translation only, so a rotating capture blurs at the limb. Measured gains,
limits and the verification against the original are in docs/derotation.md.
Feedback is genuinely wanted, especially from people who image planets:
- a capture that fails, stalls or produces a worse result than PIPP (SER/AVI plus the command line);
- workflows that are missing or awkward compared with how you actually process;
- anything the port does differently from the original that isn't listed in docs/status.md.
Open an issue, or start a discussion. "It didn't work on my data" is a useful report — attach the log
from <output>/logs/.
Parts of this codebase were written with AI assistance (large language models working from decompiled output, Ghidra listings and black-box measurements of the original), and AI will continue to be used for the modernization work. Every behaviour claim is backed by a test or a parity case against the original binaries; where something could not be verified it is recorded as a known gap in docs/status.md rather than asserted.
OPIPP is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License, version 3 or later. See LICENSE.
The licence covers this repository's own code and artwork. It does not relicense PIPP itself, and it does not make this an official release — see NOTICE.
OPIPP is an independent project. It is not affiliated with, endorsed by, sponsored by, or supported by Chris Garry or the PIPP project, and it is not an official PIPP release. "PIPP" names the original software; "OPIPP" names this independent continuation. Bug reports about the original belong with the original author, not here.
This project is a clean, managed reimplementation built for interoperability, research and
modernization: it reads the same formats, accepts the same command-line vocabulary, and reproduces the
original's behaviour so existing workflows keep working. Everything in src/ is new, readable C# —
none of the original binary is redistributed — and the project is published in the belief that this
is fair use. The decompiled and Ghidra output under decompiled/ and ghidra/ are development
reference artefacts; they are never shipped, never committed, and are git-ignored.
Provenance details, including what was and was not copied, are in notes/provenance.md.
| Artifact | Role |
|---|---|
PIPP.exe |
WinForms front-end: builds a command line, spawns the engine, parses progress |
pipp_core.exe |
the engine: decodes AVI/SER/FITS/images/RAW, debayers, calibrates, aligns, scores, stacks, writes |
pipp_dll.dll |
probe + calibration-master stacking helper (statically links libpng) |
av*.dll, sw*.dll |
FFmpeg 3.x |
The GUI is not the product — the switch vocabulary is. All 152 argument emissions the GUI produces are documented in notes/cli-spec.md.
src/Opipp.Core managed engine: formats, processing, quality, alignment, stacking, pipeline
src/Opipp.Cli drop-in replacement for pipp_core.exe
src/Opipp.App .NET 10 WinForms front-end ported from the decompiled UI
tests/ xUnit suites per subsystem + end-to-end pipeline tests
tools/ parity harness against the original engine + fixture generators
docs/ roadmap to feature parity, current status, parity guide, switch matrix
notes/ research trail: CLI spec, algorithms, engine behaviour, provenance
bin/ original binaries, copied read-only (reference only, git-ignored)
decompiled/ ILSpy project for PIPP.exe (14,277 lines of C#, reference only, git-ignored)
ghidra/ 4,229 decompiled native functions + CSV index + string-xref tooling (git-ignored)
Resuming work: docs/HANDOFF.md · progress and what comes next: docs/roadmap.md · current numbers: docs/status.md · switch-by-switch state: docs/switch-matrix.csv.
Requires the .NET 10 SDK (dotnet --version ≥ 10.0).
dotnet build OPIPP.slnx
dotnet test OPIPP.slnx
# headless engine (same switches as pipp_core.exe)
dotnet run --project src/Opipp.Cli -- -help
dotnet run --project src/Opipp.Cli -- -tif -q=50 -od=out capture.ser
# WinForms front-end
dotnet run --project src/Opipp.AppWorking end to end (verified by tests): SER → pipeline → FITS/TIFF/SER/AVI/GIF output, calibration
subtract/divide, histogram stretch, quality selection, alignment, optional field-rotation correction, stacking and animation. 1446 tests
across nine projects (0 failing; two skip without the evaluation capture); docs/status.md is the authoritative summary.
The CLI is a drop-in engine: it accepts the GUI's stdin protocol (switches and file paths, one per
line) as well as a normal command line, and emits the original @PROGRESS/@PFILE/... tokens with
-gui. Opipp.App now prefers Opipp.Cli.exe via EngineLocator and only falls back to the legacy
pipp_core.exe when the managed engine is absent.
| Area | State |
|---|---|
| SER v3 read/write | complete |
| FITS read/write | complete (8- and 16-bit, Bayer keywords, BZERO); every image unit is 2880-byte aligned |
| AVI | read: uncompressed DIB/Y800 and baseline MJPEG; write: uncompressed DIB/Y800 only. Other compressed codecs (ULRG, FMP4, YUY2, ...) are rejected with a clear error |
| BMP / TIFF / PNG | complete; BMP writes 8/24bpp only (PIPP's reader rejects 16bpp), so 16-bit frames are downshifted and documented |
| PGM / DNG | PGM is a calibration-master input only, not a science format; DNG is read-only (restricted uncompressed subset), not a general camera-RAW reader |
| Debayer (bilinear, smooth hue) | complete; pattern indices match the engine; debayered frames carry no CFA marker, an explicit pattern overrides headers, and raw geometry retags the phase |
| Binning, median, mono modes, stretch, gain/gamma, flip/rotate/crop/resize | complete |
| Dark/flat calibration | complete; masters via .tmp (preferred) or PGM, corrected flat applied exactly once |
| Quality sort/limit | complete contract; score formulas are reconstructions |
| Alignment | centroid + NCC translation + AFB box |
| Final-image stacking | -stack=mean|sum|median (OPIPP extension) and a GUI dialog; subpixel registration, coverage masks, double-precision intermediate, exact median (in memory below the caps, disk-backed above), CFA phase validated after registration |
| CLI | all 131 switch labels parse; 76 verified against the engine, 43 wired, 12 inert (see docs/status.md) |
| Not equivalent yet | surface-feature tracking (23/30 frames exact), RAW/dcraw absolute values, 16-bit Bayer debayer — see docs/status.md |
| WinForms UI | builds on .NET 10; drives the managed CLI through the original stdin/stdout protocol (legacy binary as fallback) and builds master darks/flats in-process when pipp_dll.dll is absent |
tools/Opipp.Parity runs bin/pipp_core.exe and Opipp.Cli.exe on the same synthetic capture with the
same switches and compares the produced frames by SHA-256 of their sample data:
dotnet run --project tools/Opipp.Parity
Current result: 27 of 27 compared cases match, with two deliberate differences declared separately (29 cases in total), comparing pixel content, directory layout (capture/1.tif,
logs/capture_pipp_log.txt) and log text as well as pixels — plain TIFF, -hs=204 stretch,
-flip_x, -fits, -png, -avi, -db=2 debayer, -ctr centring and -q=2 quality selection.
Quality selection is compared as a multiset because it may reorder frames, and its score column is
excluded until the quality algorithm is parity-proven.
The two declared differences in the parity set are 16-bit TIFF (the original byte-swaps samples
under an II header; we write spec-correct files) and GIF (the original emits an image-less file; we
write a valid animation). The -gui stdout is byte-identical across all nine recorded transcript
scenarios. Colour SER output is a documented superset of the original's mono-only help text, not a
parity claim (see notes/algorithms.md).
Details of what is observed versus reconstructed: notes/algorithms.md.
- notes/00-plan.md — scope, architecture, modernization decisions
- notes/engine-behaviour.md — pipeline order, constants, stdout protocol
- notes/cli-spec.md — every switch with the GUI guard that emits it
- notes/algorithms.md — evidence per algorithm
- notes/provenance.md — licensing status
