Command line¶
delta ships a standalone command-line interface (a Typer
app with Rich output). It reads FITS inputs, runs the
difference-imaging pipeline, and writes a multi-extension FITS product with provenance
header cards.
Install the cli extra to pull in Typer and Rich:
The console-script entry point is delta. Progress and per-stage timings are logged to
stderr via loguru; control verbosity with -v
(debug) / -q (warnings only). Run delta --help for the top-level help and
delta <command> --help for a full option list.
delta subtract¶
Difference two FITS images: fit the matching kernel, subtract, and write products.
delta subtract science.fits reference.fits -o diff.fits \
--gain 1.5 --read-noise 4 --decorrelate --score -v
The output is a multi-extension FITS (difference + variance/mask, plus score when
--score). With --compress (the default) the layers are RICE tile-compressed and the
difference is moved to an extension.
Arguments
| argument | meaning |
|---|---|
SCIENCE |
science image FITS path |
REFERENCE |
reference (template) image FITS path |
Options
Noise model
| option | meaning | default |
|---|---|---|
--science-var PATH |
FITS variance map for science | — |
--reference-var PATH |
FITS variance map for reference | — |
--gain FLOAT |
detector gain (e⁻/ADU); synthesises variance if no map | — |
--read-noise FLOAT |
read noise (ADU) for synthesised variance | 0.0 |
Kernel / spatial model
| option | meaning | default |
|---|---|---|
--n-max INT |
Gauss-Hermite basis order | 6 |
--beta FLOAT |
kernel basis scale (px); auto from FWHMs if omitted | — |
--n-knots INT |
thin-plate knots per axis | 5 |
--stamp-radius INT |
stamp half-size (px) | 15 |
--threshold-sigma FLOAT |
detection threshold for stamp selection | 5.0 |
--max-stamps INT |
cap on stamps used | 200 |
--saturation FLOAT |
saturation level; reject stamps above (0 = off) |
0.0 |
--catalog PATH |
text file of x y stamp centres (overrides detection) |
— |
--direction [auto\|science\|reference] |
which image to convolve (auto = the sharper) |
auto |
Post-processing
| option | meaning | default |
|---|---|---|
--decorrelate / --no-decorrelate |
ZOGY-style noise decorrelation | off |
--score / --no-score |
emit a matched-filter S/N score image | off |
--block INT |
FFT block size for decorrelation | 256 |
Output / feedback
| option | meaning | default |
|---|---|---|
-o, --output PATH |
output multi-extension FITS path (required) | — |
--save-solution PATH |
also write the kernel solution to a .npz file |
— |
--overwrite |
overwrite the output if it exists | off |
--compress / --no-compress |
RICE tile-compress the output layers | on |
-v, --verbose |
increase log verbosity (-v = debug) |
0 |
-q, --quiet |
only warnings and errors | off |
Note
The CLI exposes the common knobs; advanced Subtractor options
(bright_mask, lambda_grid, clip_sigma, clip_iterations, min_stamps,
cv_folds, spatial_scale, radius) are available only through the
Python API.
delta apply¶
Apply a saved kernel solution (.npz, from delta subtract --save-solution or
KernelSolution.save) to a new science/reference pair and
subtract — without re-fitting. The convolution direction, basis and coefficients all
come from the solution; the inputs must match its frame shape.
delta subtract sci1.fits ref.fits -o diff1.fits --save-solution kernel.npz --gain 1.5
delta apply kernel.npz sci2.fits ref.fits -o diff2.fits --gain 1.5
Arguments
| argument | meaning |
|---|---|
SOLUTION |
saved kernel solution (.npz) from a previous fit |
SCIENCE |
science image FITS path |
REFERENCE |
reference (template) image FITS path |
Options
The noise-model (--science-var, --reference-var, --gain, --read-noise),
post-processing (--decorrelate, --score, --block), and output
(-o/--output, --overwrite, --compress, -v, -q) options match
delta subtract. The kernel/spatial model is fixed by the
solution, so only --saturation (output mask level) and --stamp-radius (score
PSF half-size) remain from that group.
Provenance¶
Every FITS product written by delta subtract / delta apply (and
DiffResult.write in the Python API) carries a full
provenance header — see Provenance headers for the
complete card reference.
delta info¶
Report the data/variance/mask layers and basic statistics of a FITS file.
Prints a table of each layer's shape, dtype, and min/median/max, plus the masked-pixel fraction when a mask is present.