noobase 0.0.4

Foundational pure-function utilities for astronomy analysis

noobase

Foundational pure-function utilities for astronomy analysis. Rust core with Python bindings via PyO3.

Status: pre-1.0, API unstable. Breaking changes expected between minor versions.

Install

pip install noobase

Requires Python 3.12 or newer. Wheels are published for linux-x86_64, macos-arm64, and windows-x86_64; a source distribution is also available.

Quick start

import numpy as np
import noobase

wavelength = np.linspace(1.0, 5.0, 200)
flux = np.exp(-((wavelength - 3.0) ** 2) / 0.5)
error = 0.01 * np.ones_like(flux)

spectrum = noobase.spectroscopy.Spectrum(
    wavelength=wavelength,
    flux=flux,
    error=error,
    spacing="linear",
    kind="centers",
)

transmission_grid = np.linspace(2.5, 3.5, 50)
transmission_values = np.exp(-((transmission_grid - 3.0) ** 2) / 0.05)

band_flux, band_error, coverage = spectrum.synthetic_photometry(
    transmission_grid=transmission_grid,
    transmission_values=transmission_values,
    convention="photon_counting",
)

What's in the box

• axis.Grid — 1-D monotonic axis (linear / log, centers / edges)
• axis.overlap.{rebin, rebin_variance, coverage} — overlap-weighted rebin primitives between two axis grids
• convolve.{gaussian1d, conv1d, conv_axis, conv2d, conv2d_renorm, conv_axis_renorm} — bare correlation kernels plus NaN-as-missing renormalized variants
• spectroscopy.Spectrum — wavelength + flux + optional error + optional mask, with rebinning, flux density convention conversion, and noise-free-template LSF broadening via Spectrum.convolve_lsf
• spectroscopy.synthetic_photometry.{synthetic, SyntheticOperator} — synthetic photometry through transmission curves; the cached operator is suited for MCMC hot loops
• image.reproject_exact — surface-brightness-conserving image reprojection via planar polygon clipping (rayon-parallel; WCS handling stays in the caller's astropy / gwcs)
• image.make_pixel_corners — turn a pair of pixel_to_world_values / world_to_pixel_values callables (astropy.wcs or gwcs) into the corner array consumed by reproject_exact, with optional coarse_step for expensive WCS chains
• image.convolve_psf — true 2-D PSF convolution with NaN-as-missing edge / mask renormalization
• image.convolve_gaussian_axis — 1-D Gaussian axis correlation for grism-style matched filtering
• image.build_stamp — recenter a point-source cutout into a fixed stamp (sub-pixel centroid recorded, not applied)
• image.psf.build_epsf — oversampled ePSF from a stack of under-sampled stamps via projected Landweber / Irani–Peleg super-resolution
• image.psf.build_extended_psf — bright-star wing stacking + core↔wing stitch into an encircled-energy-normalised extended PSF, with robust_combine / solve_flux_background / stitch_psf exposed as leaves
• aperture.grow_mask — adaptive aperture mask grown by a heap-driven greedy loop from one or more seed pixels, terminated by an inner-annulus SNR stop and a radial-gradient stop (each with independent hysteresis); accepts an optional segmentation label_map + label_allowed whitelist to keep the mask inside the source's pre-computed segmentation

Convolution examples

template = noobase.spectroscopy.Spectrum(
    wavelength=wavelength,
    flux=flux,
    spacing="linear",
    kind="centers",
)
broadened = template.convolve_lsf(spec="constant_r", resolving_power=3000.0)

psf = psf / psf.sum()
model_image = noobase.image.convolve_psf(image, psf)

line_response = noobase.image.convolve_gaussian_axis(
    image,
    sigma=2.5,
    axis=0,
    normalization="l2",
)

See the full project README on GitHub for the complete feature list, the workspace layout, and the development workflow.

License

MIT.