Source code for galpak.hyperspectral_cube

# -*- coding: utf-8 -*-
from __future__ import division

import logging
import math


import numpy as np

from astropy.io import fits

logging.basicConfig(level=logging.INFO)
logHC = logging.getLogger('GalPak: HyperCube')


# Some todos
#
#Sanitization operations :
#    - IDL-made FITS :
#        - keyword FSCALE (http://astro.uni-tuebingen.de/software/idl/astrolib/fits/mrdfits.html)
#        - other non-ISO header blunders -_-
#    - CDEL_3 => CDELT3
#
#Use SI units, http://www.aanda.org/articles/aa/full_html/2010/16/aa15362-10/aa15362-10.html#S39
#Uniformization operations :
#    - Spatial Units !
#    - Spectral Units !

[docs]class HyperspectralCube: """ A hyperspectral cube has 3 dimensions : 2 spatial, and 1 spectral. It is basically a rectangular region of the sky seen on a discrete subset of the light spectrum. It is heavily used in spectral imaging, and is provided by spectroscopic explorers such as MUSE. This class is essentially a convenience wrapper for data values and header specs of a single hyperspectral cube. It understands the basic arithmetic operations ``+ - * / **``, which should behave much like with numpy's ndarrays, and update the header accordingly. An operation between a cube and a number will behave as expected, by applying the operation voxel-by-voxel to the cube. (it broadcasts the number to the shape of the cube) An operation between a cube and an image (a slice along the spectral axis, z) will behave as expected, by applying the operation image-by-image to the cube. (it broadcasts the image to the shape of the cube) It also understands indexation of the form ``cube[zmin:zmax, ymin:ymax, xmin:xmax]``. See also http://en.wikipedia.org/wiki/Hyperspectral_imaging . Note : no FSCALE management is implemented yet. Factories --------- You can load from a single HDU of a FITS (Flexible Image Transport System) file : :: HyperspectralCube.from_file(filename, hdu_index=None, verbose=False) Parameters ---------- data: numpy.ndarray|None 3D numpy ndarray holding the raw data. The indices are the pixel coordinates in the cube. Note that in order to be consistent with astropy.io.fits, data is indexed in that order : λ, y, x. header: fits.Header|None http://docs.astropy.org/en/latest/io/fits/api/headers.html Note: this might become its own wrapper class in the future. verbose: boolean Set to True to log everything and anything. """ xy_step = None z_cunit = 'Undef' z_step = None z_central = None def __init__(self, data=None, header=None, verbose=False, filename=None, variance=None): self.data = data self.header = header self.verbose = verbose self.filename = filename self.var = variance if header is not None: if not isinstance(header, fits.Header): raise TypeError("Header must be a fits Header") # Sanitize self.sanitize() #initialize try: self.initialize_self_cube() except: pass
[docs] def defaults_from_instrument(self, instrument=None): """ First reads the header for xy_step (CDELT1 or CD1_1) z_step (CDELT3 or CD3_3) z_cunit (CUNIT3) crpix (CRPIX3) crval (CRVAL3). If some of these are None, get the default values from the instrument and then reset the header : xy_step: Instrument.xy_step --> self.xy_step z_step: Instrument.z_step --> self.z_step --> CDELT3 z_cunit: Instrument.z_cunit --> self.z_cunit --> CUNIT3 crpix: shape[0]/2 --> CRPIX3 crval: Instrument.z_central --> self.z_central --> CRVAL3 for CUNIT3 CRVAL3 CRPIX3 CDELT3 """ xy_step = self.get_steps()[1] z_step = self.header.get('CDELT3') z_cunit = self.header.get('CUNIT3') crpix = self.header.get('CRPIX3') crval = self.header.get('CRVAL3') if xy_step is None: self.xy_step = instrument.cube_default_xy_step if z_step is None: self.z_step = instrument.cube_default_z_step self.header.set('CDELT3', self.z_step) if 'CD3_3' in self.header.keys(): self.header.remove('CD3_3') if z_cunit is None and instrument.cube_default_cunit is not None: self.z_cunit = instrument.cube_default_cunit self.header.set('CUNIT3', self.z_cunit) if crval is None and instrument.cube_default_z_central is not None: self.z_central = instrument.cube_default_z_central self.header.set('CRVAL3', self.z_central) if crpix is None: self.header.set('CRPIX3', (self.shape[0]) / 2.)
[docs] def initialize_self_cube(self): """ Initialize steps xy_steps z_steps z_cunnit and z_central z_central requires CRPIX3 CRVAL3 CDELT3 """ steps = self.get_steps() if (steps[1] is not None) and (steps[2] is not None): if np.abs(np.abs(steps[1]) - np.abs(steps[2])) > 1e-2: # when the size of pixels differ by more than 1% raise NotImplementedError( "GalPak does not support unsquare pixels") xy_step = steps[1] if (self.header.get('CUNIT1') is not None): if ('deg' in self.header.get('CUNIT1') ): xy_step *= 3600 # convert from deg to " if ('mas' in self.header.get('CUNIT1') ): xy_step /= 1e3 # convert from mas to " else: raise ValueError("The Cube has no cunit1 for the spaxels, " "use cunit1 to provide it") self.xy_step = xy_step self.z_step = steps[0] if self.header.get('CUNIT3') is not None: self.z_cunit = self.header['CUNIT3'] else: raise ValueError("The Cube has no cunit3 for the spaxels, " "use cunit3 to provide it") if self.z_central is None: self.z_central = self.wavelength_of(math.floor(self.shape[0] / 2.)-1)
[docs] @staticmethod def from_file(filename, hdu_index=None, verbose=False): """ Factory to create a HyperspectralCube from one HDU in a FITS file. http://fits.gsfc.nasa.gov/fits_standard.html No other file formats are supported at the moment. You may specify the index of the HDU you want. If you don't, it will try to guess by searching for a HDU with EXTNAME=DATA, or it will pick the first HDU in the list that has 3 dimensions. filename: string An absolute or relative path to the FITS file we want to load. The `astropy.io.fits` module is used to open the file. hdu_index: integer|None Index of the HDU to load. If you set this, guessing is not attempted. verbose: boolean Should we fill up the log with endless chatter ? :rtype: HyperspectralCube """ try: hdu_list = fits.open(filename) except Exception: raise IOError("Could not open provided FITS file '%s'", filename) if verbose: logHC.disabled=False else: logHC.disabled=True logHC.info("Opening %s.", filename) # Find out which HDU we should read the data from hdu = header = None extnames = [h.header.get('EXTNAME','').lower() for h in hdu_list] def _sanity_check_dimensions(header, hdu_index): if header.get('NAXIS') != 3: logHC.error("HDU dimensions : NAXIS is only #%d not 3", header.get('NAXIS')) raise ValueError("Not a cube") # 1/ If set, use HDU indexed at hdu_index if hdu_index is not None: if hdu_index < 0 or hdu_index >= len(hdu_list): raise IOError("Specified HDU #%d is not available", hdu_index) logHC.warning("Found extention %d for data" % (hdu_index)) hdu = hdu_list[hdu_index] header = hdu.header data = hdu.data if hdu_index < len(hdu_list): logHC.warning("Using next extention %d for Variance" % (hdu_index+1)) variance = hdu_list[hdu_index+1].data else: variance = None _sanity_check_dimensions(header, hdu_index) else: # 2/ Try to find DATA extension HDU found_data_hdu = False hdu_index = 0 while not found_data_hdu and hdu_index < len(hdu_list): hdu = hdu_list[hdu_index] header = hdu.header if 'data' in header.get('EXTNAME', '').lower(): try: _sanity_check_dimensions(header, hdu_index) except IOError: hdu_index += 1 else: found_data_hdu = True else: hdu_index += 1 # 3/ Try any other HDU, starting with Primary if not found_data_hdu: hdu_index = 0 while not found_data_hdu and hdu_index < len(hdu_list): hdu = hdu_list[hdu_index] header = hdu.header try: _sanity_check_dimensions(header, hdu_index) except IOError: hdu_index += 1 else: found_data_hdu = True # 4/ Alert! Exception raised if not found_data_hdu: raise IOError("Could not find a HDU containing 3 dimensional " "data in file '%s'." % filename) else: # Zealous sanity check (skipped if python is run with -O or -OO) assert hdu is not None and header is not None # Collect and numpyfy data data = np.array(hdu.data, dtype=float) #look for variance found_variance_hdu = False hdu_index = 0 if ('variance' in extnames): hdu_index = hdu_list.index_of('STAT') hdu = hdu_list[hdu_index] header = hdu.header found_variance_hdu = True if ('stat' in extnames): hdu_index = hdu_list.index_of('STAT') hdu = hdu_list[hdu_index] header = hdu.header found_variance_hdu = True if found_variance_hdu: logHC.info('Found Variance data in file %s' % (filename)) try: _sanity_check_dimensions(header, hdu_index) variance = np.array(hdu.data, dtype=float) except IOError: raise IOError("Could not find a variance with proper dimensions") else: logHC.debug("Variance data not present in file '%s'." % filename) variance = None return HyperspectralCube(data=data, header=header, verbose=verbose, filename=filename, variance=variance)
@staticmethod def from_mpdaf(cube, verbose=False): """ Factory to create a cube from a `MPDAF Cube <http://urania1.univ-lyon1.fr/mpdaf/chrome/site/DocCoreLib/user_manual_cube.html>`_. """ header = cube.data_header data = cube.data variance = cube.var return HyperspectralCube(data=data, header=header, verbose=verbose, filename=cube.filename, variance=variance) @property def shape(self): """ Returns the shape of the data, that is a tuple of the cardinality of each dimension. """ if self.data is None: return () # behave consistently with numpy.ndarray return self.data.shape
[docs] def is_empty(self): """ Is this cube void of any data ? return boolean """ return self.data is None
[docs] def has_header(self): """ Does this cube have a header ? return boolean """ return self.header is not None
def patch(self, crval3=None, crpix3=None, cunit3=None, cdelt3=None, ctype3=None, cunit1=None, force=False): """ Patches up the Cube's missing header values. Will raise a ValueError if the values are already in the header unless you use the force=. crval3: float A value for the cube's header's CRVAL3. crpix3: float A value for the cube's header's CRPIX3. cunit3: str A value for the cube's header's CUNIT3. cdelt3: float A value for the cube's header's CDELT3. ctype3: float A value for the cube's header's CTYPE3. cunit1: str A value for the cube's header's CUNIT1. force: bool Set to True if you want to overwrite existing header cards. """ logHC.info('Patching the header with cards provided') if not self.has_header(): self.header = fits.Header() if crval3 is not None: if force or self.header.get('CRVAL3') is None: self.header.set('CRVAL3', crval3) else: raise ValueError("The cube already has a CRVAL3 header card. Use the force=.") if crpix3 is not None: if force or self.header.get('CRPIX3') is None: self.header.set('CRPIX3', crpix3) else: raise ValueError("The cube already has a CRPIX3 header card. Use the force=.") if cdelt3 is not None: if force or self.header.get('CDELT3') is None: self.header.set('CDELT3', cdelt3) if 'CD3_3' in self.header.keys(): self.header.remove('CD3_3') else: raise ValueError("The cube already has a CDELT3 header card. Use the force=.") if cunit3 is not None: if force or self.header.get('CUNIT3') is None: self.header.set('CUNIT3', cunit3) else: raise ValueError("The cube already has a CUNIT3 header card. Use the force=.") if ctype3 is not None: if force or self.header.get('CTYPE3') is None: self.header.set('CTYPE3', ctype3) else: raise ValueError("The cube already has a CTYPE3 header card. Use the force=.") if cunit1 is not None: if force or self.header.get('CUNIT1') is None: self.header.set('CUNIT1', cunit1) else: raise ValueError("The cube already has a CUNIT1 header card. Use the force=.") #logging.debug('Header after patch ' + self.__str__() ) def __str__(self): """ Simple but useful string representation of the header and the data's shape. """ if self.has_header(): header = self.header.tostring("\n ").strip() else: header = "None" if self.is_empty(): data = "None" else: data = "ndarray of shape %s" % str(self.data.shape) return """HyperspectralCube header : {header} data : {data} """.format(header=header, data=data) def __add__(self, other): """ Addition should work pretty much like numpy's ndarray addition. HyperspectralCube + Number = HyperspectralCube => add the number to each voxel of the cube HyperspectralCube + HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand + on empty HyperspectralCube.") if isinstance(other, HyperspectralCube): if other.is_empty(): raise TypeError("Cannot use operand + on empty HyperspectralCube.") data = self.data + other.data else: data = self.data + other # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __radd__(self, other): """ Addition following Peano axioms is commutative on our sets. """ return self.__add__(other) def __sub__(self, other): """ Subtraction should work pretty much like numpy's ndarray subtraction. HyperspectralCube - Number = HyperspectralCube => subtract the number to each voxel of the cube HyperspectralCube - HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand - on empty HyperspectralCube.") if isinstance(other, HyperspectralCube): if other.is_empty(): raise TypeError("Cannot use operand - on empty HyperspectralCube.") data = self.data - other.data else: data = self.data - other # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __rsub__(self, other): """ Subtraction should work pretty much like numpy's ndarray subtraction. Number - HyperspectralCube = HyperspectralCube => subtract each voxel of the cube to the number, in each voxel HyperspectralCube - HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand - on empty HyperspectralCube.") if isinstance(other, HyperspectralCube): if other.is_empty(): raise TypeError("Cannot use operand - on empty HyperspectralCube.") data = other.data - self.data else: data = other - self.data # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __mul__(self, other): """ Multiplication should work pretty much like numpy's ndarray multiplication. HyperspectralCube * Number = HyperspectralCube => multiply each voxel of the cube by the number HyperspectralCube * HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand * on empty HyperspectralCube.") if isinstance(other, HyperspectralCube): if other.is_empty(): raise TypeError("Cannot use operand * on empty HyperspectralCube.") data = self.data * other.data else: data = self.data * other # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __rmul__(self, other): """ Multiplication following Peano axioms is commutative on our sets. """ return self.__mul__(other) def __truediv__(self, other): """ Division should work pretty much like numpy's ndarray division. We don't check for division by 0, and let exceptions bubble from numpy. It's faster, tends to be IEEE 754 compliant, and is consistent with numpy. TODO: find out the potential impact of from __future__ import division on this HyperspectralCube / Number = HyperspectralCube => divide each voxel of the cube by the number HyperspectralCube / HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand / on empty HyperspectralCube.") if isinstance(other, HyperspectralCube): if other.is_empty(): raise TypeError("Cannot use operand / on empty HyperspectralCube.") data = self.data / other.data else: data = self.data / other # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __rdiv__(self, other): """ Division should work pretty much like numpy's ndarray division. We don't check for division by 0, and let exceptions bubble from numpy. It's faster, tends to be IEEE 754 compliant, and is consistent with numpy. Number / HyperspectralCube = HyperspectralCube => divide each voxel of the cube by the number HyperspectralCube / HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand / on empty HyperspectralCube.") if isinstance(other, HyperspectralCube): if other.is_empty(): raise TypeError("Cannot use operand / on empty HyperspectralCube.") data = other.data / self.data else: data = other / self.data # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __pow__(self, power): """ Exponentiation should work pretty much like numpy's ndarray exponentiation. HyperspectralCube ** Number = HyperspectralCube => power each voxel of the cube by the number HyperspectralCube ** HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand ** on empty HyperspectralCube.") if isinstance(power, HyperspectralCube): if power.is_empty(): raise TypeError("Cannot use operand ** on empty HyperspectralCube.") data = self.data ** power.data else: data = self.data ** power # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __rpow__(self, base): """ Exponentiation should work pretty much like numpy's ndarray exponentiation. Number ** HyperspectralCube = HyperspectralCube => each voxel of the cube is the exponent of the number HyperspectralCube ** HyperspectralCube = HyperspectralCube => requires the input cubes to be of broadcast-compatible shapes Raises TypeErrors when cubes are empty or operands are not compatible. """ if self.is_empty(): raise TypeError("Cannot use operand ** on empty HyperspectralCube.") if isinstance(base, HyperspectralCube): if base.is_empty(): raise TypeError("Cannot use operand ** on empty HyperspectralCube.") data = base.data ** self.data else: data = base ** self.data # numpy will raise when inappropriate return HyperspectralCube(data=data, header=self.header) def __getitem__(self, item): """ HyperspectralCube[λmin:λmax, ymin:ymax, xmin:xmax] => HyperspectralCube The selector is quite similar to numpy's ndarray selector : min is included, max is excluded. You can use negative values, Python slicing syntax, as described here : http://docs.scipy.org/doc/numpy/reference/arrays.indexing.html :rtype: HyperspectralCube """ if self.is_empty(): raise TypeError("Cannot use [:,:,:] selector on empty HyperspectralCube.") data = self.data[item] if self.has_header(): header = self.header.copy() self.check_header() # todo : adjust header values to the cut raise NotImplementedError("This class does not update the headers. " "Use mpdaf.obj.Cube to cut AND update headers.") else: header = None return HyperspectralCube(data=data, header=header) def __setitem__(self, key, value): """ HyperspectralCube[λmin:λmax, ymin:ymax, xmin:xmax] = HyperspectralCube HyperspectralCube[λmin:λmax, ymin:ymax, xmin:xmax] = Number HyperspectralCube[λmin:λmax, ymin:ymax, xmin:xmax] = numpy.ndarray The indexed mutator is quite similar to numpy's ndarray's : min is included, max is excluded. You can use negative values, Python slicing syntax, as described here : http://docs.scipy.org/doc/numpy/reference/arrays.indexing.html This mutates this HyperspectralCube's data and does not return anything. value: HyperspectralCube or numpy.ndarray or Number Must be broadcastable to shape defined by the [λ,y,x] slice. """ if self.is_empty(): raise TypeError("Cannot use [:,:,:] selector on empty HyperspectralCube.") if isinstance(value, HyperspectralCube): values = value.data # fixme: make sure headers are compatible else: values = value self.data[key] = values # numpy will raise appropriate errors def copy(self, out=None): """ Copies this cube into out (if specified) and returns the copy. :rtype: HyperspectralCube """ # LBYL pattern, as self.data MUST be a numpy.ndarray and we don't want this to be silenced if self.is_empty(): data = None else: data = self.data.copy() # Same for header, MUST be an instance of astropy.io.fits.Header if self.has_header(): header = self.header.copy() else: header = None if out is None: out = HyperspectralCube(data=data, header=header) else: if not isinstance(out, HyperspectralCube): raise IOError("out parameter must be an instance of HyperspectralCube") out.data = data out.header = data return out
[docs] def write_to(self, filename, overwrite=False): """ Write this cube to a FITS file. filename: string The filename (absolute or relative) of the file we want to write to. The `astropy.io.fits` module is used to write to the file. overwrite: bool When set to True, will overwrite the output file if it exists. """ #astropy v1.2.1 IO does not work with MaskedArrays if 'ndarray' in str(type(self.data)): primary_hdu = fits.PrimaryHDU(data=self.data, header=self.header) elif 'MaskedArray' in str(type(self.data)): primary_hdu = fits.PrimaryHDU(data=self.data.data, header=self.header) else: IOError('Hyperspectral cube data is neither ndarray or MaskedArray') hdulist = fits.HDUList([primary_hdu]) hdulist.writeto(filename, overwrite=overwrite) if self.verbose: logHC.info("Writing HyperspectralCube to file %s.", filename) # Note: astropy already logs something similar, is this really necessary ? self.filename = filename
[docs] def get_steps(self): """ Returns a list of the 3 steps [λ,y,x]. The units are the ones specified in the header. """ if not self.has_header(): raise IOError("Cannot get the steps of a Cube without header.") else: #@fixme: need to fix when both present if ('CD3_3' in self.header.keys() and 'CDELT3' in self.header.keys()): logHC.info('Both CD3_3 and CDELT3 are present \n will use CDELT3 !') z_step = self.header.remove('CD3_3') z_step = self.header.get('CD3_3') or self.header.get('CDELT3') try: x_step = np.sqrt(self.header.get('CD1_1')**2. + self.header.get('CD1_2')**2. ) y_step = np.sqrt(self.header.get('CD2_2')**2. + self.header.get('CD2_1')**2. ) except: x_step = self.header.get('CDELT1') y_step = self.header.get('CDELT2') return [z_step, y_step, x_step]
def pixel_from_lambda(self, wavelength): """ Returns the pixel index (starting at 0) for the passed wavelength, whose unit must be the one specified in the header. :rtype: float """ wavelength_array = np.array(wavelength) crval = self.header.get('CRVAL3') crpix = self.header.get('CRPIX3') if 'CD3_3' in self.header.keys() and 'CDELT3' in self.header.keys(): logHC.info("CD3_3 and CDELT3 are both present\n will use CDELT3") self.header.remove('CD3_3') cdelt = self.header.get('CDELT3') or self.header.get('CD3_3') if crval is None or crpix is None or cdelt is None: raise IOError("Cube is missing required 'CRVAL3', 'CRPIX3' or 'CDELT3' header card.") return (wavelength_array - crval) / cdelt - 1.0 + crpix
[docs] def wavelength_of(self, pixel_index): """ Get the wavelength (in the unit specified in the header) of the specified pixel index along z. The pixel index should start at 0. :rtype: float """ pixel_array = np.array(pixel_index) crval = self.header.get('CRVAL3') crpix = self.header.get('CRPIX3') if 'CD3_3' in self.header.keys() and 'CDELT3' in self.header.keys(): logHC.info("CD3_3 and CDELT3 are both present\n will use CDELT3") self.header.remove('CD3_3') cdelt = self.header.get('CDELT3') or self.header.get('CD3_3') if crval is None or crpix is None or cdelt is None: raise IOError("Cube is missing required 'CRVAL3', 'CRPIX3' or 'CDELT3' header card.") return (pixel_array - crpix + 1.0) * cdelt + crval
[docs] def sanitize(self, header=None, data=None): """ Procedurally apply various sanitization tasks : - http://docs.astropy.org/en/latest/io/fits/usage/verification.html (todo) - Fix spectral step keyword : - CDEL_3 --> CDELT3 - CD3_3 --> CDELT3 - Fix blatantly illegal units : - DEG --> deg - MICRONS --> um Sanitizes this cube's header and data, or provided header and data. .. warning:: header and data are mutated, not copied, so this method returns nothing. """ if header is None: header = self.header if data is None: data = self.data if header is None: return # Fix spectral step keyword if header.get('CDELT3') is None: if header.get('CDEL_3') is not None: header.set('CDELT3', header.get('CDEL_3')) if header.get('CD3_3') is not None: header.set('CDELT3', header.get('CD3_3')) # Fix illegal units units_keywords = ['CUNIT1', 'CUNIT2', 'CUNIT3'] for unit_keyword in units_keywords: unit = header.get(unit_keyword) if unit is not None: unit = unit.strip() if unit == 'DEG': unit = 'deg' elif 'micron' in unit.lower(): unit = 'um' header.set(unit_keyword, unit)
def check_header(self): """ Will check that the header exists and has the cards we need : - CUNIT3 - CDELT3 We need those cards because they contain calibration markers (start value, step, unit) without which the data is pretty much just wasted disk space. Raises AttributeError when the header does not check out. When you load the data into a HyperspectralCube using a FITS file, the factory will require the FITS header to pass this check. (unless you specify check=False) Note that the sanitization routine will also automatically create a CDELT3 header card from a CDEL_3 header card, for example, and therefore will make this check pass. If you want this check to pass, you should add the missing cards to the header using assignations like the following (albeit after replacing values in <>) : :: self.header.set('<MISSING_CARD>', <missing_value>) """ if self.header is None: raise AttributeError("Header is not set. You need a Header for this operation.") mandatory_cards = [ 'CUNIT3' 'CDELT3' ] for card in mandatory_cards: value = self.header.get(card) if value is None: raise AttributeError("Header's '%s' card is missing, but required for this operation.", card)