Spectre Normalization Tool (SNT)
git clone [email protected]:iastro-pt/Spectra-Normalization-Tool.gitcd Spectra-Normalization-Toolpip install .See the jupyter notebook in the docs/quickstart.ipynb folder
The algorithm allows some configuration/tuning of the initial parameters.
Either do:
from SNT.utils.SNT_configs import construct_SNT_configs
confs = construct_SNT_configs()
confs.print_table_of_descriptions()Or check the full table that is provided here:
| Name | description |
|---|---|
| remove_n_first | removes the first n points in the spectra (only if first points are not useful!) otherwise set to 0 |
| radius_min | min alpha shape radius |
| radius_max | max alpha shape radius (should be at least the size of the largest gap) |
| max_vicinity | required number of values between adjacent maxima |
| stretching | normalization parameter |
| use_RIC | use RIC to avoid large 'dips' in the continuum (see documentation) |
| interp | interpolation type |
| use_denoise | for noisy spectra use the average of the flux value around the maximum |
| usefilter | use savgol filter to smooth spectra |
| nu | exponent of the computed penalty (see documentation) |
| niter_peaks_remove | number of iterations to remove sharpest peaks before interpolation |
| denoising_distance | number of points to calculate the average around a maximum if use_denoise is True, useful for noisy spectra |
Diogo Marques André M. Silva