kinextract.convolve_gaussian_pixels

kinextract.convolve_gaussian_pixels(flux: ndarray, sigma_pix: float, axis: int = -1) → ndarray[source]

Convolve flux with a Gaussian kernel of a fixed width in pixels.

Thin wrapper over scipy.ndimage.gaussian_filter1d used to apply the resolution-matching kernel from resolution_mismatch_sigma_A() (after converting its Angstrom sigma to pixels via the wavelength step). Works on a single 1-D spectrum or a 2-D template matrix (pass axis=0 for a (npix, ntemplates) matrix, matching the convention used in build_template_matrix_fortran()).

Parameters:

  • flux (ndarray) – Spectrum or template matrix to convolve.

  • sigma_pix (float) – Gaussian sigma, in pixels. If <= 0, flux is returned unchanged (no-op) rather than raising, since a zero-width match is a valid (if unusual) outcome of resolution_mismatch_sigma_A().

  • axis (int, optional) – Axis along which to convolve (the wavelength axis).

Returns:

Convolved array, same shape as flux.

Return type:

ndarray

Notes

Uses mode="nearest" boundary handling (edge values repeated) rather than wrapping or zero-padding, since spectra are not periodic and zero-padding would spuriously pull flux down near the edges.