NeutronImaging 1.6

Neutron Imaging Reduction

NeutronImagingScripts

This package contains a suite of Python modules and scripts that are critical for the data reduction of Neutron Imaging at Oak Ridge National Laboratory.

Overview

Component	What it does
neutronimaging.detector_correction	MCP detector occupancy (deadtime) correction: shutter metadata parsing, FITS stack loading with NeuNorm-1.x-equivalent gamma filtering, per-shutter-window occupancy correction, optional Timepix chip-geometry correction (gated behind --chipcorrection)
neutronimaging.preprocess	CG1D preprocessing: TIFF metadata extraction and measurement-configuration (JSON) generation
neutronimaging.npmath / neutronimaging.util	numerical and filesystem helpers (edge binning, EPICS timestamp conversion, folder trees)
mcp_detector_correction (CLI)	end-to-end MCP correction: validate input dir, correct, export float32 TIFFs under true source names + sidecar metadata
generate_config (CLI)	build the reduction configuration JSON from radiograph/OB/DF directories

Image I/O is direct astropy (FITS) and tifffile (TIFF) — the NeuNorm dependency was removed in #19.

Installation

General users

Install the package from PyPI with

$ pip install NeutronImaging

NOTE: the current PyPI release (1.2) long predates the NeuNorm removal and the 2026 correctness fixes; install from git until v1.6 is released.

Developers

Development uses pixi to manage the environment. After cloning this repository, run the following at the repo root:

$ pixi install          # solve and install the dev environment (editable package included)
$ pixi run test         # run the unit tests
$ pixi run pre-commit run --all-files   # lint

A plain-pip path also works in any Python >=3.12 environment:

$ pip install -e .

Usage

Use as a Package

Examples of using this package as a Python module are provided as Jupyter Notebooks inside the example folder.

Use as a commandline tool

Generate Configuration File for Data Reduction

To generate the json file that is needed for subsequent data reduction, use

$ generate_config IPTS-20267/raw/radiographs IPTS-20267/raw/ob IPTS-20267/raw/df IPTS-20267.json

where

• IPTS-20267/raw/radiographs contains the raw images
• IPTS-20267/raw/ob contains open beam images (white field)
• IPTS-20267/raw/df contains dark field images

If you would like to have multiple experiment configuration files nested in one json file, simply use

$ generate_config IPTS-20267/raw/radiographs,IPTS-20267-2/raw/radiographs IPTS-20267/raw/ob IPTS-20267/raw/df IPTS-20267.json

notice that:

• You can have more than one folder for raw images, but they need to be within the same string separated by ,.
• You can have only one folder for open beam directory
• You can have only one folder for dark field directory

The command above will yield a json file with the following structure

 {"IPTS-20267": {"CONFIG_DATA"},
  "IPTS-20268": {"CONFIG_DATA"}
 }

The default tolerance for the categorization with respect to aperture positions is 1mm. However, you can change the default value by specify it as below

$ generate_config \
    IPTS-20267/raw/radiographs \
    IPTS-20267/raw/ob \
    IPTS-20267/raw/df \
    IPTS-20267.json --tolerance=2

MCP Detector correction

Installing this package provides the mcp_detector_correction console command (callers that invoke the repo file by path can keep using scripts/mcp_detector_correction.py, which is a thin wrapper around the same implementation). Type mcp_detector_correction -h, you should see the following

$ mcp_detector_correction -h
Usage:
    mcp_detector_correction [--skipimg] [--verbose] <input_dir> <output_dir>
    mcp_detector_correction (-h | --help)
    mcp_detector_correction --version

Therefore, you can process the example data with the following command at the root of this repo (the data/ tree is not tracked in git — fetch it once with pixi run python -c 'from neutronimaging.datasets import ensure_repo_data; ensure_repo_data()', which downloads it into data/):

$ mcp_detector_correction data tmp

and you will see the following in your terminal

$ mcp_detector_correction data tmp
Parsing input
Validating input arguments
Processing metadata
Loading images into memory
Perform correction
corrected image summary
	dimension:	(916, 512, 512)
	type:	float64
Writing data to tmp

NOTE: make sure you create a tmp folder first.

Developer Notes

Test/example data

The ~570 MB IPTS-20267 dataset is not tracked in git. It is published as the data-v1 release asset (nis-test-data-v1.tar.gz) and fetched on demand with pooch:

• neutronimaging.datasets.example_data_dir() returns the data root, downloading it into the pooch cache on first use (resolution order: $NEUTRONIMAGING_DATA_DIR → a populated local data/ → download). The test suite uses this and skips if the data is unreachable.
• neutronimaging.datasets.ensure_repo_data() materializes it at data/ so the example notebooks' ../data relative paths resolve.
• pixi run fetch-data pre-downloads it (CI runs this before the tests).

Contents: OB_1_005_*.fits (916 frames) + OB_1_005_{ShutterCount,ShutterTimes,Spectra}.txt (an MCP open-beam measurement for the detector-correction tests/notebook); IPTS-20267/ (CG1D directory trees exercising generate_config); and the golden profiles ref_pop_prof.npz / ref_img_mean_prof.npz for the occupancy correction. Regenerate the goldens ONLY after an intentional, reviewed behavior change with pixi run python scripts/regenerate_test_refs.py, then re-tar and bump the tag + sha256 in neutronimaging/datasets.py. (No LFS — deliberately abandoned in ef81aed.)

Releasing

1. Tag vX.Y on main (versioningit derives versions from tags).
2. Build and upload manually (python -m build, twine upload dist/*) — by decision there is no publish automation for this repo (no known downstream pip consumers).

The package is MIT-licensed (team decision 2026-06-12, the NDP standard), which settled the historical GPLv3/BSD/LGPLv2 metadata contradiction.

Branch model

main only (decided 2026-06-10): work lands on main via PRs; the historical next/qa promotion branches are frozen relics of the pre-2026 flow.