Description
- Classes:
- Alias:
pixel_replace
During 1-D spectral extraction (extract_1d step), pixels flagged as bad are ignored in the summation process. If a bad pixel is part of the point-spread function (PSF) at a given wavelength, the absence of the signal in the flagged pixel will lead to a hollow space at that wavelength in the extracted spectrum.
To avoid this defect in the 1-D spectrum, this step estimates the flux values of pixels
flagged as DO_NOT_USE in 2-D extracted spectra using interpolation methods,
prior to rectification in the resample_spec step.
pixel_replace inserts these estimates into the 2-D data array,
unsets the DO_NOT_USE flag, and sets the FLUX_ESTIMATED flag for each affected pixel.
This step is provided as a cosmetic feature and, for that reason, should be used with caution.
Algorithms
Adjacent Profile Approximation
This is the default (and most extensively tested) algorithm for most spectroscopic modes.
First, the input 2-D spectral cutout is scanned across the dispersion axis to determine which cross-dispersion vectors (column or row, depending on dispersion direction) contain at least one flagged pixel. Next, for each affected vector, a median normalized profile is created.
The adjacent arrays (the number of which is set by the step argument
n_adjacent_cols) are individually normalized. Next, each pixel in the profile is set to
the median of the normalized values. This results in a median of normalized values filling the vector.
Finally, this profile is scaled to the vector containing a missing pixel, and the value is estimated from the scaled profile.
Minimum Gradient Estimator
In the case of the MIRI MRS, NaN-valued pixels are partially compensated during the IFU cube building process using the overlap between detector pixels and output cube voxels. The effects of NaN values are thus not as severe as for slit spectra, but can manifest as small dips in the extracted spectrum when a NaN value lands atop the peak of a spectral trace and cube building interpolates from lower-flux adjacent values.
Pixel replacement can thus be useful in some science cases for the MIRI MRS as well, but undersampling combined with the curvature of spectral traces on the detector can lead the model-based adjacent profile estimator to derive incorrect values in the vicinity of emission lines. The minimum gradient estimator is thus another optional algorithm that uses entirely local information to fill in the missing pixel values.
This method tests the gradient along the spatial and spectral axes using immediately adjacent pixels. It chooses whichever dimension has the minimum absolute gradient and replaces the missing pixel with the average of the two adjacent pixels along that dimension. Near point sources this will thus favor replacement along the spectral axis due to spatial undersampling of the PSF profile, while near bright extended emission lines it will favor replacement along the spatial axis due to the steep spectral profile. No replacement is attempted if a NaN value is bordered by another NaN value along a given axis.