lsst.pipe.tasks  19.0.0-45-g25df2313
dcrAssembleCoadd.py
Go to the documentation of this file.
1 # This file is part of pipe_tasks.
2 #
3 # LSST Data Management System
4 # This product includes software developed by the
5 # LSST Project (http://www.lsst.org/).
6 # See COPYRIGHT file at the top of the source tree.
7 #
8 # This program is free software: you can redistribute it and/or modify
9 # it under the terms of the GNU General Public License as published by
10 # the Free Software Foundation, either version 3 of the License, or
11 # (at your option) any later version.
12 #
13 # This program is distributed in the hope that it will be useful,
14 # but WITHOUT ANY WARRANTY; without even the implied warranty of
15 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 # GNU General Public License for more details.
17 #
18 # You should have received a copy of the LSST License Statement and
19 # the GNU General Public License along with this program. If not,
20 # see <https://www.lsstcorp.org/LegalNotices/>.
21 #
22 
23 from math import ceil
24 import numpy as np
25 from scipy import ndimage
26 import lsst.geom as geom
27 import lsst.afw.image as afwImage
28 import lsst.afw.table as afwTable
29 import lsst.coadd.utils as coaddUtils
30 from lsst.daf.butler import DeferredDatasetHandle
31 from lsst.ip.diffim.dcrModel import applyDcr, calculateDcr, DcrModel
32 import lsst.meas.algorithms as measAlg
33 from lsst.meas.base import SingleFrameMeasurementTask
34 import lsst.pex.config as pexConfig
35 import lsst.pipe.base as pipeBase
36 import lsst.utils as utils
37 from .assembleCoadd import (AssembleCoaddTask,
38  CompareWarpAssembleCoaddConfig,
39  CompareWarpAssembleCoaddTask)
40 from .coaddBase import makeSkyInfo
41 from .measurePsf import MeasurePsfTask
42 
43 __all__ = ["DcrAssembleCoaddConnections", "DcrAssembleCoaddTask", "DcrAssembleCoaddConfig"]
44 
45 
46 class DcrAssembleCoaddConnections(pipeBase.PipelineTaskConnections,
47  dimensions=("tract", "patch", "abstract_filter", "skymap"),
48  defaultTemplates={"inputCoaddName": "deep",
49  "outputCoaddName": "dcr",
50  "warpType": "direct",
51  "warpTypeSuffix": "",
52  "fakesType": ""}):
53  inputWarps = pipeBase.connectionTypes.Input(
54  doc=("Input list of warps to be assembled i.e. stacked."
55  "WarpType (e.g. direct, psfMatched) is controlled by the warpType config parameter"),
56  name="{inputCoaddName}Coadd_{warpType}Warp",
57  storageClass="ExposureF",
58  dimensions=("tract", "patch", "skymap", "visit", "instrument"),
59  deferLoad=True,
60  multiple=True
61  )
62  skyMap = pipeBase.connectionTypes.Input(
63  doc="Input definition of geometry/bbox and projection/wcs for coadded exposures",
64  name="{inputCoaddName}Coadd_skyMap",
65  storageClass="SkyMap",
66  dimensions=("skymap", ),
67  )
68  brightObjectMask = pipeBase.connectionTypes.PrerequisiteInput(
69  doc=("Input Bright Object Mask mask produced with external catalogs to be applied to the mask plane"
70  " BRIGHT_OBJECT."),
71  name="brightObjectMask",
72  storageClass="ObjectMaskCatalog",
73  dimensions=("tract", "patch", "skymap", "abstract_filter"),
74  )
75  templateExposure = pipeBase.connectionTypes.Input(
76  doc="Input coadded exposure, produced by previous call to AssembleCoadd",
77  name="{fakesType}{inputCoaddName}Coadd{warpTypeSuffix}",
78  storageClass="ExposureF",
79  dimensions=("tract", "patch", "skymap", "abstract_filter"),
80  )
81  dcrCoadds = pipeBase.connectionTypes.Output(
82  doc="Output coadded exposure, produced by stacking input warps",
83  name="{fakesType}{outputCoaddName}Coadd{warpTypeSuffix}",
84  storageClass="ExposureF",
85  dimensions=("tract", "patch", "skymap", "abstract_filter", "subfilter"),
86  multiple=True,
87  )
88  dcrNImages = pipeBase.connectionTypes.Output(
89  doc="Output image of number of input images per pixel",
90  name="{outputCoaddName}Coadd_nImage",
91  storageClass="ImageU",
92  dimensions=("tract", "patch", "skymap", "abstract_filter", "subfilter"),
93  multiple=True,
94  )
95 
96  def __init__(self, *, config=None):
97  super().__init__(config=config)
98  if not config.doWrite:
99  self.outputs.remove("dcrCoadds")
100 
101 
102 class DcrAssembleCoaddConfig(CompareWarpAssembleCoaddConfig,
103  pipelineConnections=DcrAssembleCoaddConnections):
104  dcrNumSubfilters = pexConfig.Field(
105  dtype=int,
106  doc="Number of sub-filters to forward model chromatic effects to fit the supplied exposures.",
107  default=3,
108  )
109  maxNumIter = pexConfig.Field(
110  dtype=int,
111  optional=True,
112  doc="Maximum number of iterations of forward modeling.",
113  default=None,
114  )
115  minNumIter = pexConfig.Field(
116  dtype=int,
117  optional=True,
118  doc="Minimum number of iterations of forward modeling.",
119  default=None,
120  )
121  convergenceThreshold = pexConfig.Field(
122  dtype=float,
123  doc="Target relative change in convergence between iterations of forward modeling.",
124  default=0.001,
125  )
126  useConvergence = pexConfig.Field(
127  dtype=bool,
128  doc="Use convergence test as a forward modeling end condition?"
129  "If not set, skips calculating convergence and runs for ``maxNumIter`` iterations",
130  default=True,
131  )
132  baseGain = pexConfig.Field(
133  dtype=float,
134  optional=True,
135  doc="Relative weight to give the new solution vs. the last solution when updating the model."
136  "A value of 1.0 gives equal weight to both solutions."
137  "Small values imply slower convergence of the solution, but can "
138  "help prevent overshooting and failures in the fit."
139  "If ``baseGain`` is None, a conservative gain "
140  "will be calculated from the number of subfilters. ",
141  default=None,
142  )
143  useProgressiveGain = pexConfig.Field(
144  dtype=bool,
145  doc="Use a gain that slowly increases above ``baseGain`` to accelerate convergence? "
146  "When calculating the next gain, we use up to 5 previous gains and convergence values."
147  "Can be set to False to force the model to change at the rate of ``baseGain``. ",
148  default=True,
149  )
150  doAirmassWeight = pexConfig.Field(
151  dtype=bool,
152  doc="Weight exposures by airmass? Useful if there are relatively few high-airmass observations.",
153  default=False,
154  )
155  modelWeightsWidth = pexConfig.Field(
156  dtype=float,
157  doc="Width of the region around detected sources to include in the DcrModel.",
158  default=3,
159  )
160  useModelWeights = pexConfig.Field(
161  dtype=bool,
162  doc="Width of the region around detected sources to include in the DcrModel.",
163  default=True,
164  )
165  splitSubfilters = pexConfig.Field(
166  dtype=bool,
167  doc="Calculate DCR for two evenly-spaced wavelengths in each subfilter."
168  "Instead of at the midpoint",
169  default=True,
170  )
171  splitThreshold = pexConfig.Field(
172  dtype=float,
173  doc="Minimum DCR difference within a subfilter to use ``splitSubfilters``, in pixels."
174  "Set to 0 to always split the subfilters.",
175  default=0.1,
176  )
177  regularizeModelIterations = pexConfig.Field(
178  dtype=float,
179  doc="Maximum relative change of the model allowed between iterations."
180  "Set to zero to disable.",
181  default=2.,
182  )
183  regularizeModelFrequency = pexConfig.Field(
184  dtype=float,
185  doc="Maximum relative change of the model allowed between subfilters."
186  "Set to zero to disable.",
187  default=4.,
188  )
189  convergenceMaskPlanes = pexConfig.ListField(
190  dtype=str,
191  default=["DETECTED"],
192  doc="Mask planes to use to calculate convergence."
193  )
194  regularizationWidth = pexConfig.Field(
195  dtype=int,
196  default=2,
197  doc="Minimum radius of a region to include in regularization, in pixels."
198  )
199  imageInterpOrder = pexConfig.Field(
200  dtype=int,
201  doc="The order of the spline interpolation used to shift the image plane.",
202  default=3,
203  )
204  accelerateModel = pexConfig.Field(
205  dtype=float,
206  doc="Factor to amplify the differences between model planes by to speed convergence.",
207  default=3,
208  )
209  doCalculatePsf = pexConfig.Field(
210  dtype=bool,
211  doc="Set to detect stars and recalculate the PSF from the final coadd."
212  "Otherwise the PSF is estimated from a selection of the best input exposures",
213  default=False,
214  )
215  detectPsfSources = pexConfig.ConfigurableField(
216  target=measAlg.SourceDetectionTask,
217  doc="Task to detect sources for PSF measurement, if ``doCalculatePsf`` is set.",
218  )
219  measurePsfSources = pexConfig.ConfigurableField(
220  target=SingleFrameMeasurementTask,
221  doc="Task to measure sources for PSF measurement, if ``doCalculatePsf`` is set."
222  )
223  measurePsf = pexConfig.ConfigurableField(
224  target=MeasurePsfTask,
225  doc="Task to measure the PSF of the coadd, if ``doCalculatePsf`` is set.",
226  )
227 
228  def setDefaults(self):
229  CompareWarpAssembleCoaddConfig.setDefaults(self)
230  self.assembleStaticSkyModel.retarget(CompareWarpAssembleCoaddTask)
231  self.doNImage = True
232  self.assembleStaticSkyModel.warpType = self.warpType
233  # The deepCoadd and nImage files will be overwritten by this Task, so don't write them the first time
234  self.assembleStaticSkyModel.doNImage = False
235  self.assembleStaticSkyModel.doWrite = False
236  self.detectPsfSources.returnOriginalFootprints = False
237  self.detectPsfSources.thresholdPolarity = "positive"
238  # Only use bright sources for PSF measurement
239  self.detectPsfSources.thresholdValue = 50
240  self.detectPsfSources.nSigmaToGrow = 2
241  # A valid star for PSF measurement should at least fill 5x5 pixels
242  self.detectPsfSources.minPixels = 25
243  # Use the variance plane to calculate signal to noise
244  self.detectPsfSources.thresholdType = "pixel_stdev"
245  # The signal to noise limit is good enough, while the flux limit is set
246  # in dimensionless units and may not be appropriate for all data sets.
247  self.measurePsf.starSelector["objectSize"].doFluxLimit = False
248 
249 
250 class DcrAssembleCoaddTask(CompareWarpAssembleCoaddTask):
251  """Assemble DCR coadded images from a set of warps.
252 
253  Attributes
254  ----------
255  bufferSize : `int`
256  The number of pixels to grow each subregion by to allow for DCR.
257 
258  Notes
259  -----
260  As with AssembleCoaddTask, we want to assemble a coadded image from a set of
261  Warps (also called coadded temporary exposures), including the effects of
262  Differential Chromatic Refraction (DCR).
263  For full details of the mathematics and algorithm, please see
264  DMTN-037: DCR-matched template generation (https://dmtn-037.lsst.io).
265 
266  This Task produces a DCR-corrected deepCoadd, as well as a dcrCoadd for
267  each subfilter used in the iterative calculation.
268  It begins by dividing the bandpass-defining filter into N equal bandwidth
269  "subfilters", and divides the flux in each pixel from an initial coadd
270  equally into each as a "dcrModel". Because the airmass and parallactic
271  angle of each individual exposure is known, we can calculate the shift
272  relative to the center of the band in each subfilter due to DCR. For each
273  exposure we apply this shift as a linear transformation to the dcrModels
274  and stack the results to produce a DCR-matched exposure. The matched
275  exposures are subtracted from the input exposures to produce a set of
276  residual images, and these residuals are reverse shifted for each
277  exposures' subfilters and stacked. The shifted and stacked residuals are
278  added to the dcrModels to produce a new estimate of the flux in each pixel
279  within each subfilter. The dcrModels are solved for iteratively, which
280  continues until the solution from a new iteration improves by less than
281  a set percentage, or a maximum number of iterations is reached.
282  Two forms of regularization are employed to reduce unphysical results.
283  First, the new solution is averaged with the solution from the previous
284  iteration, which mitigates oscillating solutions where the model
285  overshoots with alternating very high and low values.
286  Second, a common degeneracy when the data have a limited range of airmass or
287  parallactic angle values is for one subfilter to be fit with very low or
288  negative values, while another subfilter is fit with very high values. This
289  typically appears in the form of holes next to sources in one subfilter,
290  and corresponding extended wings in another. Because each subfilter has
291  a narrow bandwidth we assume that physical sources that are above the noise
292  level will not vary in flux by more than a factor of `frequencyClampFactor`
293  between subfilters, and pixels that have flux deviations larger than that
294  factor will have the excess flux distributed evenly among all subfilters.
295  If `splitSubfilters` is set, then each subfilter will be further sub-
296  divided during the forward modeling step (only). This approximates using
297  a higher number of subfilters that may be necessary for high airmass
298  observations, but does not increase the number of free parameters in the
299  fit. This is needed when there are high airmass observations which would
300  otherwise have significant DCR even within a subfilter. Because calculating
301  the shifted images takes most of the time, splitting the subfilters is
302  turned off by way of the `splitThreshold` option for low-airmass
303  observations that do not suffer from DCR within a subfilter.
304  """
305 
306  ConfigClass = DcrAssembleCoaddConfig
307  _DefaultName = "dcrAssembleCoadd"
308 
309  def __init__(self, *args, **kwargs):
310  super().__init__(*args, **kwargs)
311  if self.config.doCalculatePsf:
312  self.schema = afwTable.SourceTable.makeMinimalSchema()
313  self.makeSubtask("detectPsfSources", schema=self.schema)
314  self.makeSubtask("measurePsfSources", schema=self.schema)
315  self.makeSubtask("measurePsf", schema=self.schema)
316 
317  @utils.inheritDoc(pipeBase.PipelineTask)
318  def runQuantum(self, butlerQC, inputRefs, outputRefs):
319  # Docstring to be formatted with info from PipelineTask.runQuantum
320  """
321  Notes
322  -----
323  Assemble a coadd from a set of Warps.
324 
325  PipelineTask (Gen3) entry point to Coadd a set of Warps.
326  Analogous to `runDataRef`, it prepares all the data products to be
327  passed to `run`, and processes the results before returning a struct
328  of results to be written out. AssembleCoadd cannot fit all Warps in memory.
329  Therefore, its inputs are accessed subregion by subregion
330  by the Gen3 `DeferredDatasetHandle` that is analagous to the Gen2
331  `lsst.daf.persistence.ButlerDataRef`. Any updates to this method should
332  correspond to an update in `runDataRef` while both entry points
333  are used.
334  """
335  inputData = butlerQC.get(inputRefs)
336 
337  # Construct skyInfo expected by run
338  # Do not remove skyMap from inputData in case makeSupplementaryDataGen3 needs it
339  skyMap = inputData["skyMap"]
340  outputDataId = butlerQC.quantum.dataId
341 
342  inputData['skyInfo'] = makeSkyInfo(skyMap,
343  tractId=outputDataId['tract'],
344  patchId=outputDataId['patch'])
345 
346  # Construct list of input Deferred Datasets
347  # These quack a bit like like Gen2 DataRefs
348  warpRefList = inputData['inputWarps']
349  # Perform same middle steps as `runDataRef` does
350  inputs = self.prepareInputs(warpRefList)
351  self.log.info("Found %d %s", len(inputs.tempExpRefList),
352  self.getTempExpDatasetName(self.warpType))
353  if len(inputs.tempExpRefList) == 0:
354  self.log.warn("No coadd temporary exposures found")
355  return
356 
357  supplementaryData = self.makeSupplementaryDataGen3(butlerQC, inputRefs, outputRefs)
358  retStruct = self.run(inputData['skyInfo'], inputs.tempExpRefList, inputs.imageScalerList,
359  inputs.weightList, supplementaryData=supplementaryData)
360 
361  inputData.setdefault('brightObjectMask', None)
362  for subfilter in range(self.config.dcrNumSubfilters):
363  # Use the PSF of the stacked dcrModel, and do not recalculate the PSF for each subfilter
364  retStruct.dcrCoadds[subfilter].setPsf(retStruct.coaddExposure.getPsf())
365  self.processResults(retStruct.dcrCoadds[subfilter], inputData['brightObjectMask'], outputDataId)
366 
367  if self.config.doWrite:
368  butlerQC.put(retStruct, outputRefs)
369  return retStruct
370 
371  @pipeBase.timeMethod
372  def runDataRef(self, dataRef, selectDataList=None, warpRefList=None):
373  """Assemble a coadd from a set of warps.
374 
375  Coadd a set of Warps. Compute weights to be applied to each Warp and
376  find scalings to match the photometric zeropoint to a reference Warp.
377  Assemble the Warps using run method.
378  Forward model chromatic effects across multiple subfilters,
379  and subtract from the input Warps to build sets of residuals.
380  Use the residuals to construct a new ``DcrModel`` for each subfilter,
381  and iterate until the model converges.
382  Interpolate over NaNs and optionally write the coadd to disk.
383  Return the coadded exposure.
384 
385  Parameters
386  ----------
387  dataRef : `lsst.daf.persistence.ButlerDataRef`
388  Data reference defining the patch for coaddition and the
389  reference Warp
390  selectDataList : `list` of `lsst.daf.persistence.ButlerDataRef`
391  List of data references to warps. Data to be coadded will be
392  selected from this list based on overlap with the patch defined by
393  the data reference.
394 
395  Returns
396  -------
397  results : `lsst.pipe.base.Struct`
398  The Struct contains the following fields:
399 
400  - ``coaddExposure``: coadded exposure (`lsst.afw.image.Exposure`)
401  - ``nImage``: exposure count image (`lsst.afw.image.ImageU`)
402  - ``dcrCoadds``: `list` of coadded exposures for each subfilter
403  - ``dcrNImages``: `list` of exposure count images for each subfilter
404  """
405  if (selectDataList is None and warpRefList is None) or (selectDataList and warpRefList):
406  raise RuntimeError("runDataRef must be supplied either a selectDataList or warpRefList")
407 
408  skyInfo = self.getSkyInfo(dataRef)
409  if warpRefList is None:
410  calExpRefList = self.selectExposures(dataRef, skyInfo, selectDataList=selectDataList)
411  if len(calExpRefList) == 0:
412  self.log.warn("No exposures to coadd")
413  return
414  self.log.info("Coadding %d exposures", len(calExpRefList))
415 
416  warpRefList = self.getTempExpRefList(dataRef, calExpRefList)
417 
418  inputData = self.prepareInputs(warpRefList)
419  self.log.info("Found %d %s", len(inputData.tempExpRefList),
420  self.getTempExpDatasetName(self.warpType))
421  if len(inputData.tempExpRefList) == 0:
422  self.log.warn("No coadd temporary exposures found")
423  return
424 
425  supplementaryData = self.makeSupplementaryData(dataRef, warpRefList=inputData.tempExpRefList)
426 
427  results = self.run(skyInfo, inputData.tempExpRefList, inputData.imageScalerList,
428  inputData.weightList, supplementaryData=supplementaryData)
429  if results is None:
430  self.log.warn("Could not construct DcrModel for patch %s: no data to coadd.",
431  skyInfo.patchInfo.getIndex())
432  return
433 
434  if self.config.doCalculatePsf:
435  self.measureCoaddPsf(results.coaddExposure)
436  brightObjects = self.readBrightObjectMasks(dataRef) if self.config.doMaskBrightObjects else None
437  for subfilter in range(self.config.dcrNumSubfilters):
438  # Use the PSF of the stacked dcrModel, and do not recalculate the PSF for each subfilter
439  results.dcrCoadds[subfilter].setPsf(results.coaddExposure.getPsf())
440  self.processResults(results.dcrCoadds[subfilter],
441  brightObjectMasks=brightObjects, dataId=dataRef.dataId)
442  if self.config.doWrite:
443  self.log.info("Persisting dcrCoadd")
444  dataRef.put(results.dcrCoadds[subfilter], "dcrCoadd", subfilter=subfilter,
445  numSubfilters=self.config.dcrNumSubfilters)
446  if self.config.doNImage and results.dcrNImages is not None:
447  dataRef.put(results.dcrNImages[subfilter], "dcrCoadd_nImage", subfilter=subfilter,
448  numSubfilters=self.config.dcrNumSubfilters)
449 
450  return results
451 
452  @utils.inheritDoc(AssembleCoaddTask)
453  def makeSupplementaryDataGen3(self, butlerQC, inputRefs, outputRefs):
454  """Load the previously-generated template coadd.
455 
456  This can be removed entirely once we no longer support the Gen 2 butler.
457 
458  Returns
459  -------
460  templateCoadd : `lsst.pipe.base.Struct`
461  Result struct with components:
462 
463  - ``templateCoadd``: coadded exposure (`lsst.afw.image.ExposureF`)
464  """
465  templateCoadd = butlerQC.get(inputRefs.templateExposure)
466 
467  return pipeBase.Struct(templateCoadd=templateCoadd)
468 
469  def measureCoaddPsf(self, coaddExposure):
470  """Detect sources on the coadd exposure and measure the final PSF.
471 
472  Parameters
473  ----------
474  coaddExposure : `lsst.afw.image.Exposure`
475  The final coadded exposure.
476  """
477  table = afwTable.SourceTable.make(self.schema)
478  detResults = self.detectPsfSources.run(table, coaddExposure, clearMask=False)
479  coaddSources = detResults.sources
480  self.measurePsfSources.run(
481  measCat=coaddSources,
482  exposure=coaddExposure
483  )
484  # Measure the PSF on the stacked subfilter coadds if possible.
485  # We should already have a decent estimate of the coadd PSF, however,
486  # so in case of any errors simply log them as a warning and use the
487  # default PSF.
488  try:
489  psfResults = self.measurePsf.run(coaddExposure, coaddSources)
490  except Exception as e:
491  self.log.warn("Unable to calculate PSF, using default coadd PSF: %s" % e)
492  else:
493  coaddExposure.setPsf(psfResults.psf)
494 
495  def prepareDcrInputs(self, templateCoadd, warpRefList, weightList):
496  """Prepare the DCR coadd by iterating through the visitInfo of the input warps.
497 
498  Sets the property ``bufferSize``.
499 
500  Parameters
501  ----------
502  templateCoadd : `lsst.afw.image.ExposureF`
503  The initial coadd exposure before accounting for DCR.
504  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
505  `lsst.daf.persistence.ButlerDataRef`
506  The data references to the input warped exposures.
507  weightList : `list` of `float`
508  The weight to give each input exposure in the coadd
509  Will be modified in place if ``doAirmassWeight`` is set.
510 
511  Returns
512  -------
513  dcrModels : `lsst.pipe.tasks.DcrModel`
514  Best fit model of the true sky after correcting chromatic effects.
515 
516  Raises
517  ------
518  NotImplementedError
519  If ``lambdaMin`` is missing from the Mapper class of the obs package being used.
520  """
521  sigma2fwhm = 2.*np.sqrt(2.*np.log(2.))
522  filterInfo = templateCoadd.getFilter()
523  if np.isnan(filterInfo.getFilterProperty().getLambdaMin()):
524  raise NotImplementedError("No minimum/maximum wavelength information found"
525  " in the filter definition! Please add lambdaMin and lambdaMax"
526  " to the Mapper class in your obs package.")
527  tempExpName = self.getTempExpDatasetName(self.warpType)
528  dcrShifts = []
529  airmassDict = {}
530  angleDict = {}
531  psfSizeDict = {}
532  for visitNum, warpExpRef in enumerate(warpRefList):
533  if isinstance(warpExpRef, DeferredDatasetHandle):
534  # Gen 3 API
535  visitInfo = warpExpRef.get(component="visitInfo")
536  psf = warpExpRef.get(component="psf")
537  else:
538  # Gen 2 API. Delete this when Gen 2 retired
539  visitInfo = warpExpRef.get(tempExpName + "_visitInfo")
540  psf = warpExpRef.get(tempExpName).getPsf()
541  visit = warpExpRef.dataId["visit"]
542  psfSize = psf.computeShape().getDeterminantRadius()*sigma2fwhm
543  airmass = visitInfo.getBoresightAirmass()
544  parallacticAngle = visitInfo.getBoresightParAngle().asDegrees()
545  airmassDict[visit] = airmass
546  angleDict[visit] = parallacticAngle
547  psfSizeDict[visit] = psfSize
548  if self.config.doAirmassWeight:
549  weightList[visitNum] *= airmass
550  dcrShifts.append(np.max(np.abs(calculateDcr(visitInfo, templateCoadd.getWcs(),
551  filterInfo, self.config.dcrNumSubfilters))))
552  self.log.info("Selected airmasses:\n%s", airmassDict)
553  self.log.info("Selected parallactic angles:\n%s", angleDict)
554  self.log.info("Selected PSF sizes:\n%s", psfSizeDict)
555  self.bufferSize = int(np.ceil(np.max(dcrShifts)) + 1)
556  psf = self.selectCoaddPsf(templateCoadd, warpRefList)
557  dcrModels = DcrModel.fromImage(templateCoadd.maskedImage,
558  self.config.dcrNumSubfilters,
559  filterInfo=filterInfo,
560  psf=psf)
561  return dcrModels
562 
563  def run(self, skyInfo, warpRefList, imageScalerList, weightList,
564  supplementaryData=None):
565  """Assemble the coadd.
566 
567  Requires additional inputs Struct ``supplementaryData`` to contain a
568  ``templateCoadd`` that serves as the model of the static sky.
569 
570  Find artifacts and apply them to the warps' masks creating a list of
571  alternative masks with a new "CLIPPED" plane and updated "NO_DATA" plane
572  Then pass these alternative masks to the base class's assemble method.
573 
574  Divide the ``templateCoadd`` evenly between each subfilter of a
575  ``DcrModel`` as the starting best estimate of the true wavelength-
576  dependent sky. Forward model the ``DcrModel`` using the known
577  chromatic effects in each subfilter and calculate a convergence metric
578  based on how well the modeled template matches the input warps. If
579  the convergence has not yet reached the desired threshold, then shift
580  and stack the residual images to build a new ``DcrModel``. Apply
581  conditioning to prevent oscillating solutions between iterations or
582  between subfilters.
583 
584  Once the ``DcrModel`` reaches convergence or the maximum number of
585  iterations has been reached, fill the metadata for each subfilter
586  image and make them proper ``coaddExposure``s.
587 
588  Parameters
589  ----------
590  skyInfo : `lsst.pipe.base.Struct`
591  Patch geometry information, from getSkyInfo
592  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
593  `lsst.daf.persistence.ButlerDataRef`
594  The data references to the input warped exposures.
595  imageScalerList : `list` of `lsst.pipe.task.ImageScaler`
596  The image scalars correct for the zero point of the exposures.
597  weightList : `list` of `float`
598  The weight to give each input exposure in the coadd
599  supplementaryData : `lsst.pipe.base.Struct`
600  Result struct returned by ``makeSupplementaryData`` with components:
601 
602  - ``templateCoadd``: coadded exposure (`lsst.afw.image.Exposure`)
603 
604  Returns
605  -------
606  result : `lsst.pipe.base.Struct`
607  Result struct with components:
608 
609  - ``coaddExposure``: coadded exposure (`lsst.afw.image.Exposure`)
610  - ``nImage``: exposure count image (`lsst.afw.image.ImageU`)
611  - ``dcrCoadds``: `list` of coadded exposures for each subfilter
612  - ``dcrNImages``: `list` of exposure count images for each subfilter
613  """
614  minNumIter = self.config.minNumIter or self.config.dcrNumSubfilters
615  maxNumIter = self.config.maxNumIter or self.config.dcrNumSubfilters*3
616  templateCoadd = supplementaryData.templateCoadd
617  baseMask = templateCoadd.mask.clone()
618  # The variance plane is for each subfilter
619  # and should be proportionately lower than the full-band image
620  baseVariance = templateCoadd.variance.clone()
621  baseVariance /= self.config.dcrNumSubfilters
622  spanSetMaskList = self.findArtifacts(templateCoadd, warpRefList, imageScalerList)
623  # Note that the mask gets cleared in ``findArtifacts``, but we want to preserve the mask.
624  templateCoadd.setMask(baseMask)
625  badMaskPlanes = self.config.badMaskPlanes[:]
626  # Note that is important that we do not add "CLIPPED" to ``badMaskPlanes``
627  # This is because pixels in observations that are significantly affect by DCR
628  # are likely to have many pixels that are both "DETECTED" and "CLIPPED",
629  # but those are necessary to constrain the DCR model.
630  badPixelMask = templateCoadd.mask.getPlaneBitMask(badMaskPlanes)
631 
632  stats = self.prepareStats(mask=badPixelMask)
633  dcrModels = self.prepareDcrInputs(templateCoadd, warpRefList, weightList)
634  if self.config.doNImage:
635  dcrNImages, dcrWeights = self.calculateNImage(dcrModels, skyInfo.bbox, warpRefList,
636  spanSetMaskList, stats.ctrl)
637  nImage = afwImage.ImageU(skyInfo.bbox)
638  # Note that this nImage will be a factor of dcrNumSubfilters higher than
639  # the nImage returned by assembleCoadd for most pixels. This is because each
640  # subfilter may have a different nImage, and fractional values are not allowed.
641  for dcrNImage in dcrNImages:
642  nImage += dcrNImage
643  else:
644  dcrNImages = None
645 
646  subregionSize = geom.Extent2I(*self.config.subregionSize)
647  nSubregions = (ceil(skyInfo.bbox.getHeight()/subregionSize[1]) *
648  ceil(skyInfo.bbox.getWidth()/subregionSize[0]))
649  subIter = 0
650  for subBBox in self._subBBoxIter(skyInfo.bbox, subregionSize):
651  modelIter = 0
652  subIter += 1
653  self.log.info("Computing coadd over patch %s subregion %s of %s: %s",
654  skyInfo.patchInfo.getIndex(), subIter, nSubregions, subBBox)
655  dcrBBox = geom.Box2I(subBBox)
656  dcrBBox.grow(self.bufferSize)
657  dcrBBox.clip(dcrModels.bbox)
658  modelWeights = self.calculateModelWeights(dcrModels, dcrBBox)
659  subExposures = self.loadSubExposures(dcrBBox, stats.ctrl, warpRefList,
660  imageScalerList, spanSetMaskList)
661  convergenceMetric = self.calculateConvergence(dcrModels, subExposures, subBBox,
662  warpRefList, weightList, stats.ctrl)
663  self.log.info("Initial convergence : %s", convergenceMetric)
664  convergenceList = [convergenceMetric]
665  gainList = []
666  convergenceCheck = 1.
667  refImage = templateCoadd.image
668  while (convergenceCheck > self.config.convergenceThreshold or modelIter <= minNumIter):
669  gain = self.calculateGain(convergenceList, gainList)
670  self.dcrAssembleSubregion(dcrModels, subExposures, subBBox, dcrBBox, warpRefList,
671  stats.ctrl, convergenceMetric, gain,
672  modelWeights, refImage, dcrWeights)
673  if self.config.useConvergence:
674  convergenceMetric = self.calculateConvergence(dcrModels, subExposures, subBBox,
675  warpRefList, weightList, stats.ctrl)
676  if convergenceMetric == 0:
677  self.log.warn("Coadd patch %s subregion %s had convergence metric of 0.0 which is "
678  "most likely due to there being no valid data in the region.",
679  skyInfo.patchInfo.getIndex(), subIter)
680  break
681  convergenceCheck = (convergenceList[-1] - convergenceMetric)/convergenceMetric
682  if (convergenceCheck < 0) & (modelIter > minNumIter):
683  self.log.warn("Coadd patch %s subregion %s diverged before reaching maximum "
684  "iterations or desired convergence improvement of %s."
685  " Divergence: %s",
686  skyInfo.patchInfo.getIndex(), subIter,
687  self.config.convergenceThreshold, convergenceCheck)
688  break
689  convergenceList.append(convergenceMetric)
690  if modelIter > maxNumIter:
691  if self.config.useConvergence:
692  self.log.warn("Coadd patch %s subregion %s reached maximum iterations "
693  "before reaching desired convergence improvement of %s."
694  " Final convergence improvement: %s",
695  skyInfo.patchInfo.getIndex(), subIter,
696  self.config.convergenceThreshold, convergenceCheck)
697  break
698 
699  if self.config.useConvergence:
700  self.log.info("Iteration %s with convergence metric %s, %.4f%% improvement (gain: %.2f)",
701  modelIter, convergenceMetric, 100.*convergenceCheck, gain)
702  modelIter += 1
703  else:
704  if self.config.useConvergence:
705  self.log.info("Coadd patch %s subregion %s finished with "
706  "convergence metric %s after %s iterations",
707  skyInfo.patchInfo.getIndex(), subIter, convergenceMetric, modelIter)
708  else:
709  self.log.info("Coadd patch %s subregion %s finished after %s iterations",
710  skyInfo.patchInfo.getIndex(), subIter, modelIter)
711  if self.config.useConvergence and convergenceMetric > 0:
712  self.log.info("Final convergence improvement was %.4f%% overall",
713  100*(convergenceList[0] - convergenceMetric)/convergenceMetric)
714 
715  dcrCoadds = self.fillCoadd(dcrModels, skyInfo, warpRefList, weightList,
716  calibration=self.scaleZeroPoint.getPhotoCalib(),
717  coaddInputs=templateCoadd.getInfo().getCoaddInputs(),
718  mask=baseMask,
719  variance=baseVariance)
720  coaddExposure = self.stackCoadd(dcrCoadds)
721  return pipeBase.Struct(coaddExposure=coaddExposure, nImage=nImage,
722  dcrCoadds=dcrCoadds, dcrNImages=dcrNImages)
723 
724  def calculateNImage(self, dcrModels, bbox, warpRefList, spanSetMaskList, statsCtrl):
725  """Calculate the number of exposures contributing to each subfilter.
726 
727  Parameters
728  ----------
729  dcrModels : `lsst.pipe.tasks.DcrModel`
730  Best fit model of the true sky after correcting chromatic effects.
731  bbox : `lsst.geom.box.Box2I`
732  Bounding box of the patch to coadd.
733  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
734  `lsst.daf.persistence.ButlerDataRef`
735  The data references to the input warped exposures.
736  spanSetMaskList : `list` of `dict` containing spanSet lists, or None
737  Each element of the `dict` contains the new mask plane name
738  (e.g. "CLIPPED and/or "NO_DATA") as the key,
739  and the list of SpanSets to apply to the mask.
740  statsCtrl : `lsst.afw.math.StatisticsControl`
741  Statistics control object for coadd
742 
743  Returns
744  -------
745  dcrNImages : `list` of `lsst.afw.image.ImageU`
746  List of exposure count images for each subfilter
747  dcrWeights : `list` of `lsst.afw.image.ImageF`
748  Per-pixel weights for each subfilter.
749  Equal to 1/(number of unmasked images contributing to each pixel).
750  """
751  dcrNImages = [afwImage.ImageU(bbox) for subfilter in range(self.config.dcrNumSubfilters)]
752  dcrWeights = [afwImage.ImageF(bbox) for subfilter in range(self.config.dcrNumSubfilters)]
753  tempExpName = self.getTempExpDatasetName(self.warpType)
754  for warpExpRef, altMaskSpans in zip(warpRefList, spanSetMaskList):
755  if isinstance(warpExpRef, DeferredDatasetHandle):
756  # Gen 3 API
757  exposure = warpExpRef.get(parameters={'bbox': bbox})
758  else:
759  # Gen 2 API. Delete this when Gen 2 retired
760  exposure = warpExpRef.get(tempExpName + "_sub", bbox=bbox)
761  visitInfo = exposure.getInfo().getVisitInfo()
762  wcs = exposure.getInfo().getWcs()
763  mask = exposure.mask
764  if altMaskSpans is not None:
765  self.applyAltMaskPlanes(mask, altMaskSpans)
766  weightImage = np.zeros_like(exposure.image.array)
767  weightImage[(mask.array & statsCtrl.getAndMask()) == 0] = 1.
768  # The weights must be shifted in exactly the same way as the residuals,
769  # because they will be used as the denominator in the weighted average of residuals.
770  weightsGenerator = self.dcrResiduals(weightImage, visitInfo, wcs, dcrModels.filter)
771  for shiftedWeights, dcrNImage, dcrWeight in zip(weightsGenerator, dcrNImages, dcrWeights):
772  dcrNImage.array += np.rint(shiftedWeights).astype(dcrNImage.array.dtype)
773  dcrWeight.array += shiftedWeights
774  # Exclude any pixels that don't have at least one exposure contributing in all subfilters
775  weightsThreshold = 1.
776  goodPix = dcrWeights[0].array > weightsThreshold
777  for weights in dcrWeights[1:]:
778  goodPix = (weights.array > weightsThreshold) & goodPix
779  for subfilter in range(self.config.dcrNumSubfilters):
780  dcrWeights[subfilter].array[goodPix] = 1./dcrWeights[subfilter].array[goodPix]
781  dcrWeights[subfilter].array[~goodPix] = 0.
782  dcrNImages[subfilter].array[~goodPix] = 0
783  return (dcrNImages, dcrWeights)
784 
785  def dcrAssembleSubregion(self, dcrModels, subExposures, bbox, dcrBBox, warpRefList,
786  statsCtrl, convergenceMetric,
787  gain, modelWeights, refImage, dcrWeights):
788  """Assemble the DCR coadd for a sub-region.
789 
790  Build a DCR-matched template for each input exposure, then shift the
791  residuals according to the DCR in each subfilter.
792  Stack the shifted residuals and apply them as a correction to the
793  solution from the previous iteration.
794  Restrict the new model solutions from varying by more than a factor of
795  `modelClampFactor` from the last solution, and additionally restrict the
796  individual subfilter models from varying by more than a factor of
797  `frequencyClampFactor` from their average.
798  Finally, mitigate potentially oscillating solutions by averaging the new
799  solution with the solution from the previous iteration, weighted by
800  their convergence metric.
801 
802  Parameters
803  ----------
804  dcrModels : `lsst.pipe.tasks.DcrModel`
805  Best fit model of the true sky after correcting chromatic effects.
806  subExposures : `dict` of `lsst.afw.image.ExposureF`
807  The pre-loaded exposures for the current subregion.
808  bbox : `lsst.geom.box.Box2I`
809  Bounding box of the subregion to coadd.
810  dcrBBox : `lsst.geom.box.Box2I`
811  Sub-region of the coadd which includes a buffer to allow for DCR.
812  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
813  `lsst.daf.persistence.ButlerDataRef`
814  The data references to the input warped exposures.
815  statsCtrl : `lsst.afw.math.StatisticsControl`
816  Statistics control object for coadd
817  convergenceMetric : `float`
818  Quality of fit metric for the matched templates of the input images.
819  gain : `float`, optional
820  Relative weight to give the new solution when updating the model.
821  modelWeights : `numpy.ndarray` or `float`
822  A 2D array of weight values that tapers smoothly to zero away from detected sources.
823  Set to a placeholder value of 1.0 if ``self.config.useModelWeights`` is False.
824  refImage : `lsst.afw.image.Image`
825  A reference image used to supply the default pixel values.
826  dcrWeights : `list` of `lsst.afw.image.Image`
827  Per-pixel weights for each subfilter.
828  Equal to 1/(number of unmasked images contributing to each pixel).
829  """
830  residualGeneratorList = []
831 
832  for warpExpRef in warpRefList:
833  visit = warpExpRef.dataId["visit"]
834  exposure = subExposures[visit]
835  visitInfo = exposure.getInfo().getVisitInfo()
836  wcs = exposure.getInfo().getWcs()
837  templateImage = dcrModels.buildMatchedTemplate(exposure=exposure,
838  order=self.config.imageInterpOrder,
839  splitSubfilters=self.config.splitSubfilters,
840  splitThreshold=self.config.splitThreshold,
841  amplifyModel=self.config.accelerateModel)
842  residual = exposure.image.array - templateImage.array
843  # Note that the variance plane here is used to store weights, not the actual variance
844  residual *= exposure.variance.array
845  # The residuals are stored as a list of generators.
846  # This allows the residual for a given subfilter and exposure to be created
847  # on the fly, instead of needing to store them all in memory.
848  residualGeneratorList.append(self.dcrResiduals(residual, visitInfo, wcs, dcrModels.filter))
849 
850  dcrSubModelOut = self.newModelFromResidual(dcrModels, residualGeneratorList, dcrBBox, statsCtrl,
851  gain=gain,
852  modelWeights=modelWeights,
853  refImage=refImage,
854  dcrWeights=dcrWeights)
855  dcrModels.assign(dcrSubModelOut, bbox)
856 
857  def dcrResiduals(self, residual, visitInfo, wcs, filterInfo):
858  """Prepare a residual image for stacking in each subfilter by applying the reverse DCR shifts.
859 
860  Parameters
861  ----------
862  residual : `numpy.ndarray`
863  The residual masked image for one exposure,
864  after subtracting the matched template
865  visitInfo : `lsst.afw.image.VisitInfo`
866  Metadata for the exposure.
867  wcs : `lsst.afw.geom.SkyWcs`
868  Coordinate system definition (wcs) for the exposure.
869  filterInfo : `lsst.afw.image.Filter`
870  The filter definition, set in the current instruments' obs package.
871  Required for any calculation of DCR, including making matched templates.
872 
873  Yields
874  ------
875  residualImage : `numpy.ndarray`
876  The residual image for the next subfilter, shifted for DCR.
877  """
878  # Pre-calculate the spline-filtered residual image, so that step can be
879  # skipped in the shift calculation in `applyDcr`.
880  filteredResidual = ndimage.spline_filter(residual, order=self.config.imageInterpOrder)
881  # Note that `splitSubfilters` is always turned off in the reverse direction.
882  # This option introduces additional blurring if applied to the residuals.
883  dcrShift = calculateDcr(visitInfo, wcs, filterInfo, self.config.dcrNumSubfilters,
884  splitSubfilters=False)
885  for dcr in dcrShift:
886  yield applyDcr(filteredResidual, dcr, useInverse=True, splitSubfilters=False,
887  doPrefilter=False, order=self.config.imageInterpOrder)
888 
889  def newModelFromResidual(self, dcrModels, residualGeneratorList, dcrBBox, statsCtrl,
890  gain, modelWeights, refImage, dcrWeights):
891  """Calculate a new DcrModel from a set of image residuals.
892 
893  Parameters
894  ----------
895  dcrModels : `lsst.pipe.tasks.DcrModel`
896  Current model of the true sky after correcting chromatic effects.
897  residualGeneratorList : `generator` of `numpy.ndarray`
898  The residual image for the next subfilter, shifted for DCR.
899  dcrBBox : `lsst.geom.box.Box2I`
900  Sub-region of the coadd which includes a buffer to allow for DCR.
901  statsCtrl : `lsst.afw.math.StatisticsControl`
902  Statistics control object for coadd
903  gain : `float`
904  Relative weight to give the new solution when updating the model.
905  modelWeights : `numpy.ndarray` or `float`
906  A 2D array of weight values that tapers smoothly to zero away from detected sources.
907  Set to a placeholder value of 1.0 if ``self.config.useModelWeights`` is False.
908  refImage : `lsst.afw.image.Image`
909  A reference image used to supply the default pixel values.
910  dcrWeights : `list` of `lsst.afw.image.Image`
911  Per-pixel weights for each subfilter.
912  Equal to 1/(number of unmasked images contributing to each pixel).
913 
914  Returns
915  -------
916  dcrModel : `lsst.pipe.tasks.DcrModel`
917  New model of the true sky after correcting chromatic effects.
918  """
919  newModelImages = []
920  for subfilter, model in enumerate(dcrModels):
921  residualsList = [next(residualGenerator) for residualGenerator in residualGeneratorList]
922  residual = np.sum(residualsList, axis=0)
923  residual *= dcrWeights[subfilter][dcrBBox].array
924  # `MaskedImage`s only support in-place addition, so rename for readability
925  newModel = model[dcrBBox].clone()
926  newModel.array += residual
927  # Catch any invalid values
928  badPixels = ~np.isfinite(newModel.array)
929  newModel.array[badPixels] = model[dcrBBox].array[badPixels]
930  if self.config.regularizeModelIterations > 0:
931  dcrModels.regularizeModelIter(subfilter, newModel, dcrBBox,
932  self.config.regularizeModelIterations,
933  self.config.regularizationWidth)
934  newModelImages.append(newModel)
935  if self.config.regularizeModelFrequency > 0:
936  dcrModels.regularizeModelFreq(newModelImages, dcrBBox, statsCtrl,
937  self.config.regularizeModelFrequency,
938  self.config.regularizationWidth)
939  dcrModels.conditionDcrModel(newModelImages, dcrBBox, gain=gain)
940  self.applyModelWeights(newModelImages, refImage[dcrBBox], modelWeights)
941  return DcrModel(newModelImages, dcrModels.filter, dcrModels.psf,
942  dcrModels.mask, dcrModels.variance)
943 
944  def calculateConvergence(self, dcrModels, subExposures, bbox, warpRefList, weightList, statsCtrl):
945  """Calculate a quality of fit metric for the matched templates.
946 
947  Parameters
948  ----------
949  dcrModels : `lsst.pipe.tasks.DcrModel`
950  Best fit model of the true sky after correcting chromatic effects.
951  subExposures : `dict` of `lsst.afw.image.ExposureF`
952  The pre-loaded exposures for the current subregion.
953  bbox : `lsst.geom.box.Box2I`
954  Sub-region to coadd
955  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
956  `lsst.daf.persistence.ButlerDataRef`
957  The data references to the input warped exposures.
958  weightList : `list` of `float`
959  The weight to give each input exposure in the coadd
960  statsCtrl : `lsst.afw.math.StatisticsControl`
961  Statistics control object for coadd
962 
963  Returns
964  -------
965  convergenceMetric : `float`
966  Quality of fit metric for all input exposures, within the sub-region
967  """
968  significanceImage = np.abs(dcrModels.getReferenceImage(bbox))
969  nSigma = 3.
970  significanceImage += nSigma*dcrModels.calculateNoiseCutoff(dcrModels[1], statsCtrl,
971  bufferSize=self.bufferSize)
972  if np.max(significanceImage) == 0:
973  significanceImage += 1.
974  weight = 0
975  metric = 0.
976  metricList = {}
977  for warpExpRef, expWeight in zip(warpRefList, weightList):
978  visit = warpExpRef.dataId["visit"]
979  exposure = subExposures[visit][bbox]
980  singleMetric = self.calculateSingleConvergence(dcrModels, exposure, significanceImage, statsCtrl)
981  metric += singleMetric
982  metricList[visit] = singleMetric
983  weight += 1.
984  self.log.info("Individual metrics:\n%s", metricList)
985  return 1.0 if weight == 0.0 else metric/weight
986 
987  def calculateSingleConvergence(self, dcrModels, exposure, significanceImage, statsCtrl):
988  """Calculate a quality of fit metric for a single matched template.
989 
990  Parameters
991  ----------
992  dcrModels : `lsst.pipe.tasks.DcrModel`
993  Best fit model of the true sky after correcting chromatic effects.
994  exposure : `lsst.afw.image.ExposureF`
995  The input warped exposure to evaluate.
996  significanceImage : `numpy.ndarray`
997  Array of weights for each pixel corresponding to its significance
998  for the convergence calculation.
999  statsCtrl : `lsst.afw.math.StatisticsControl`
1000  Statistics control object for coadd
1001 
1002  Returns
1003  -------
1004  convergenceMetric : `float`
1005  Quality of fit metric for one exposure, within the sub-region.
1006  """
1007  convergeMask = exposure.mask.getPlaneBitMask(self.config.convergenceMaskPlanes)
1008  templateImage = dcrModels.buildMatchedTemplate(exposure=exposure,
1009  order=self.config.imageInterpOrder,
1010  splitSubfilters=self.config.splitSubfilters,
1011  splitThreshold=self.config.splitThreshold,
1012  amplifyModel=self.config.accelerateModel)
1013  diffVals = np.abs(exposure.image.array - templateImage.array)*significanceImage
1014  refVals = np.abs(exposure.image.array + templateImage.array)*significanceImage/2.
1015 
1016  finitePixels = np.isfinite(diffVals)
1017  goodMaskPixels = (exposure.mask.array & statsCtrl.getAndMask()) == 0
1018  convergeMaskPixels = exposure.mask.array & convergeMask > 0
1019  usePixels = finitePixels & goodMaskPixels & convergeMaskPixels
1020  if np.sum(usePixels) == 0:
1021  metric = 0.
1022  else:
1023  diffUse = diffVals[usePixels]
1024  refUse = refVals[usePixels]
1025  metric = np.sum(diffUse/np.median(diffUse))/np.sum(refUse/np.median(diffUse))
1026  return metric
1027 
1028  def stackCoadd(self, dcrCoadds):
1029  """Add a list of sub-band coadds together.
1030 
1031  Parameters
1032  ----------
1033  dcrCoadds : `list` of `lsst.afw.image.ExposureF`
1034  A list of coadd exposures, each exposure containing
1035  the model for one subfilter.
1036 
1037  Returns
1038  -------
1039  coaddExposure : `lsst.afw.image.ExposureF`
1040  A single coadd exposure that is the sum of the sub-bands.
1041  """
1042  coaddExposure = dcrCoadds[0].clone()
1043  for coadd in dcrCoadds[1:]:
1044  coaddExposure.maskedImage += coadd.maskedImage
1045  return coaddExposure
1046 
1047  def fillCoadd(self, dcrModels, skyInfo, warpRefList, weightList, calibration=None, coaddInputs=None,
1048  mask=None, variance=None):
1049  """Create a list of coadd exposures from a list of masked images.
1050 
1051  Parameters
1052  ----------
1053  dcrModels : `lsst.pipe.tasks.DcrModel`
1054  Best fit model of the true sky after correcting chromatic effects.
1055  skyInfo : `lsst.pipe.base.Struct`
1056  Patch geometry information, from getSkyInfo
1057  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
1058  `lsst.daf.persistence.ButlerDataRef`
1059  The data references to the input warped exposures.
1060  weightList : `list` of `float`
1061  The weight to give each input exposure in the coadd
1062  calibration : `lsst.afw.Image.PhotoCalib`, optional
1063  Scale factor to set the photometric calibration of an exposure.
1064  coaddInputs : `lsst.afw.Image.CoaddInputs`, optional
1065  A record of the observations that are included in the coadd.
1066  mask : `lsst.afw.image.Mask`, optional
1067  Optional mask to override the values in the final coadd.
1068  variance : `lsst.afw.image.Image`, optional
1069  Optional variance plane to override the values in the final coadd.
1070 
1071  Returns
1072  -------
1073  dcrCoadds : `list` of `lsst.afw.image.ExposureF`
1074  A list of coadd exposures, each exposure containing
1075  the model for one subfilter.
1076  """
1077  dcrCoadds = []
1078  refModel = dcrModels.getReferenceImage()
1079  for model in dcrModels:
1080  if self.config.accelerateModel > 1:
1081  model.array = (model.array - refModel)*self.config.accelerateModel + refModel
1082  coaddExposure = afwImage.ExposureF(skyInfo.bbox, skyInfo.wcs)
1083  if calibration is not None:
1084  coaddExposure.setPhotoCalib(calibration)
1085  if coaddInputs is not None:
1086  coaddExposure.getInfo().setCoaddInputs(coaddInputs)
1087  # Set the metadata for the coadd, including PSF and aperture corrections.
1088  self.assembleMetadata(coaddExposure, warpRefList, weightList)
1089  # Overwrite the PSF
1090  coaddExposure.setPsf(dcrModels.psf)
1091  coaddUtils.setCoaddEdgeBits(dcrModels.mask[skyInfo.bbox], dcrModels.variance[skyInfo.bbox])
1092  maskedImage = afwImage.MaskedImageF(dcrModels.bbox)
1093  maskedImage.image = model
1094  maskedImage.mask = dcrModels.mask
1095  maskedImage.variance = dcrModels.variance
1096  coaddExposure.setMaskedImage(maskedImage[skyInfo.bbox])
1097  coaddExposure.setPhotoCalib(self.scaleZeroPoint.getPhotoCalib())
1098  if mask is not None:
1099  coaddExposure.setMask(mask)
1100  if variance is not None:
1101  coaddExposure.setVariance(variance)
1102  dcrCoadds.append(coaddExposure)
1103  return dcrCoadds
1104 
1105  def calculateGain(self, convergenceList, gainList):
1106  """Calculate the gain to use for the current iteration.
1107 
1108  After calculating a new DcrModel, each value is averaged with the
1109  value in the corresponding pixel from the previous iteration. This
1110  reduces oscillating solutions that iterative techniques are plagued by,
1111  and speeds convergence. By far the biggest changes to the model
1112  happen in the first couple iterations, so we can also use a more
1113  aggressive gain later when the model is changing slowly.
1114 
1115  Parameters
1116  ----------
1117  convergenceList : `list` of `float`
1118  The quality of fit metric from each previous iteration.
1119  gainList : `list` of `float`
1120  The gains used in each previous iteration: appended with the new
1121  gain value.
1122  Gains are numbers between ``self.config.baseGain`` and 1.
1123 
1124  Returns
1125  -------
1126  gain : `float`
1127  Relative weight to give the new solution when updating the model.
1128  A value of 1.0 gives equal weight to both solutions.
1129 
1130  Raises
1131  ------
1132  ValueError
1133  If ``len(convergenceList) != len(gainList)+1``.
1134  """
1135  nIter = len(convergenceList)
1136  if nIter != len(gainList) + 1:
1137  raise ValueError("convergenceList (%d) must be one element longer than gainList (%d)."
1138  % (len(convergenceList), len(gainList)))
1139 
1140  if self.config.baseGain is None:
1141  # If ``baseGain`` is not set, calculate it from the number of DCR subfilters
1142  # The more subfilters being modeled, the lower the gain should be.
1143  baseGain = 1./(self.config.dcrNumSubfilters - 1)
1144  else:
1145  baseGain = self.config.baseGain
1146 
1147  if self.config.useProgressiveGain and nIter > 2:
1148  # To calculate the best gain to use, compare the past gains that have been used
1149  # with the resulting convergences to estimate the best gain to use.
1150  # Algorithmically, this is a Kalman filter.
1151  # If forward modeling proceeds perfectly, the convergence metric should
1152  # asymptotically approach a final value.
1153  # We can estimate that value from the measured changes in convergence
1154  # weighted by the gains used in each previous iteration.
1155  estFinalConv = [((1 + gainList[i])*convergenceList[i + 1] - convergenceList[i])/gainList[i]
1156  for i in range(nIter - 1)]
1157  # The convergence metric is strictly positive, so if the estimated final convergence is
1158  # less than zero, force it to zero.
1159  estFinalConv = np.array(estFinalConv)
1160  estFinalConv[estFinalConv < 0] = 0
1161  # Because the estimate may slowly change over time, only use the most recent measurements.
1162  estFinalConv = np.median(estFinalConv[max(nIter - 5, 0):])
1163  lastGain = gainList[-1]
1164  lastConv = convergenceList[-2]
1165  newConv = convergenceList[-1]
1166  # The predicted convergence is the value we would get if the new model calculated
1167  # in the previous iteration was perfect. Recall that the updated model that is
1168  # actually used is the gain-weighted average of the new and old model,
1169  # so the convergence would be similarly weighted.
1170  predictedConv = (estFinalConv*lastGain + lastConv)/(1. + lastGain)
1171  # If the measured and predicted convergence are very close, that indicates
1172  # that our forward model is accurate and we can use a more aggressive gain
1173  # If the measured convergence is significantly worse (or better!) than predicted,
1174  # that indicates that the model is not converging as expected and
1175  # we should use a more conservative gain.
1176  delta = (predictedConv - newConv)/((lastConv - estFinalConv)/(1 + lastGain))
1177  newGain = 1 - abs(delta)
1178  # Average the gains to prevent oscillating solutions.
1179  newGain = (newGain + lastGain)/2.
1180  gain = max(baseGain, newGain)
1181  else:
1182  gain = baseGain
1183  gainList.append(gain)
1184  return gain
1185 
1186  def calculateModelWeights(self, dcrModels, dcrBBox):
1187  """Build an array that smoothly tapers to 0 away from detected sources.
1188 
1189  Parameters
1190  ----------
1191  dcrModels : `lsst.pipe.tasks.DcrModel`
1192  Best fit model of the true sky after correcting chromatic effects.
1193  dcrBBox : `lsst.geom.box.Box2I`
1194  Sub-region of the coadd which includes a buffer to allow for DCR.
1195 
1196  Returns
1197  -------
1198  weights : `numpy.ndarray` or `float`
1199  A 2D array of weight values that tapers smoothly to zero away from detected sources.
1200  Set to a placeholder value of 1.0 if ``self.config.useModelWeights`` is False.
1201 
1202  Raises
1203  ------
1204  ValueError
1205  If ``useModelWeights`` is set and ``modelWeightsWidth`` is negative.
1206  """
1207  if not self.config.useModelWeights:
1208  return 1.0
1209  if self.config.modelWeightsWidth < 0:
1210  raise ValueError("modelWeightsWidth must not be negative if useModelWeights is set")
1211  convergeMask = dcrModels.mask.getPlaneBitMask(self.config.convergenceMaskPlanes)
1212  convergeMaskPixels = dcrModels.mask[dcrBBox].array & convergeMask > 0
1213  weights = np.zeros_like(dcrModels[0][dcrBBox].array)
1214  weights[convergeMaskPixels] = 1.
1215  weights = ndimage.filters.gaussian_filter(weights, self.config.modelWeightsWidth)
1216  weights /= np.max(weights)
1217  return weights
1218 
1219  def applyModelWeights(self, modelImages, refImage, modelWeights):
1220  """Smoothly replace model pixel values with those from a
1221  reference at locations away from detected sources.
1222 
1223  Parameters
1224  ----------
1225  modelImages : `list` of `lsst.afw.image.Image`
1226  The new DCR model images from the current iteration.
1227  The values will be modified in place.
1228  refImage : `lsst.afw.image.MaskedImage`
1229  A reference image used to supply the default pixel values.
1230  modelWeights : `numpy.ndarray` or `float`
1231  A 2D array of weight values that tapers smoothly to zero away from detected sources.
1232  Set to a placeholder value of 1.0 if ``self.config.useModelWeights`` is False.
1233  """
1234  if self.config.useModelWeights:
1235  for model in modelImages:
1236  model.array *= modelWeights
1237  model.array += refImage.array*(1. - modelWeights)/self.config.dcrNumSubfilters
1238 
1239  def loadSubExposures(self, bbox, statsCtrl, warpRefList, imageScalerList, spanSetMaskList):
1240  """Pre-load sub-regions of a list of exposures.
1241 
1242  Parameters
1243  ----------
1244  bbox : `lsst.geom.box.Box2I`
1245  Sub-region to coadd
1246  statsCtrl : `lsst.afw.math.StatisticsControl`
1247  Statistics control object for coadd
1248  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
1249  `lsst.daf.persistence.ButlerDataRef`
1250  The data references to the input warped exposures.
1251  imageScalerList : `list` of `lsst.pipe.task.ImageScaler`
1252  The image scalars correct for the zero point of the exposures.
1253  spanSetMaskList : `list` of `dict` containing spanSet lists, or None
1254  Each element is dict with keys = mask plane name to add the spans to
1255 
1256  Returns
1257  -------
1258  subExposures : `dict`
1259  The `dict` keys are the visit IDs,
1260  and the values are `lsst.afw.image.ExposureF`
1261  The pre-loaded exposures for the current subregion.
1262  The variance plane contains weights, and not the variance
1263  """
1264  tempExpName = self.getTempExpDatasetName(self.warpType)
1265  zipIterables = zip(warpRefList, imageScalerList, spanSetMaskList)
1266  subExposures = {}
1267  for warpExpRef, imageScaler, altMaskSpans in zipIterables:
1268  if isinstance(warpExpRef, DeferredDatasetHandle):
1269  exposure = warpExpRef.get(parameters={'bbox': bbox})
1270  else:
1271  exposure = warpExpRef.get(tempExpName + "_sub", bbox=bbox)
1272  visit = warpExpRef.dataId["visit"]
1273  if altMaskSpans is not None:
1274  self.applyAltMaskPlanes(exposure.mask, altMaskSpans)
1275  imageScaler.scaleMaskedImage(exposure.maskedImage)
1276  # Note that the variance plane here is used to store weights, not the actual variance
1277  exposure.variance.array[:, :] = 0.
1278  # Set the weight of unmasked pixels to 1.
1279  exposure.variance.array[(exposure.mask.array & statsCtrl.getAndMask()) == 0] = 1.
1280  # Set the image value of masked pixels to zero.
1281  # This eliminates needing the mask plane when stacking images in ``newModelFromResidual``
1282  exposure.image.array[(exposure.mask.array & statsCtrl.getAndMask()) > 0] = 0.
1283  subExposures[visit] = exposure
1284  return subExposures
1285 
1286  def selectCoaddPsf(self, templateCoadd, warpRefList):
1287  """Compute the PSF of the coadd from the exposures with the best seeing.
1288 
1289  Parameters
1290  ----------
1291  templateCoadd : `lsst.afw.image.ExposureF`
1292  The initial coadd exposure before accounting for DCR.
1293  warpRefList : `list` of `lsst.daf.butler.DeferredDatasetHandle` or
1294  `lsst.daf.persistence.ButlerDataRef`
1295  The data references to the input warped exposures.
1296 
1297  Returns
1298  -------
1299  psf : `lsst.meas.algorithms.CoaddPsf`
1300  The average PSF of the input exposures with the best seeing.
1301  """
1302  sigma2fwhm = 2.*np.sqrt(2.*np.log(2.))
1303  tempExpName = self.getTempExpDatasetName(self.warpType)
1304  # Note: ``ccds`` is a `lsst.afw.table.ExposureCatalog` with one entry per ccd and per visit
1305  # If there are multiple ccds, it will have that many times more elements than ``warpExpRef``
1306  ccds = templateCoadd.getInfo().getCoaddInputs().ccds
1307  psfRefSize = templateCoadd.getPsf().computeShape().getDeterminantRadius()*sigma2fwhm
1308  psfSizes = np.zeros(len(ccds))
1309  ccdVisits = np.array(ccds["visit"])
1310  for warpExpRef in warpRefList:
1311  if isinstance(warpExpRef, DeferredDatasetHandle):
1312  # Gen 3 API
1313  psf = warpExpRef.get(component="psf")
1314  else:
1315  # Gen 2 API. Delete this when Gen 2 retired
1316  psf = warpExpRef.get(tempExpName).getPsf()
1317  visit = warpExpRef.dataId["visit"]
1318  psfSize = psf.computeShape().getDeterminantRadius()*sigma2fwhm
1319  psfSizes[ccdVisits == visit] = psfSize
1320  # Note that the input PSFs include DCR, which should be absent from the DcrCoadd
1321  # The selected PSFs are those that have a FWHM less than or equal to the smaller
1322  # of the mean or median FWHM of the input exposures.
1323  sizeThreshold = min(np.median(psfSizes), psfRefSize)
1324  goodPsfs = psfSizes <= sizeThreshold
1325  psf = measAlg.CoaddPsf(ccds[goodPsfs], templateCoadd.getWcs(),
1326  self.config.coaddPsf.makeControl())
1327  return psf
def selectCoaddPsf(self, templateCoadd, warpRefList)
def dcrResiduals(self, residual, visitInfo, wcs, filterInfo)
def calculateSingleConvergence(self, dcrModels, exposure, significanceImage, statsCtrl)
def loadSubExposures(self, bbox, statsCtrl, warpRefList, imageScalerList, spanSetMaskList)
def dcrAssembleSubregion(self, dcrModels, subExposures, bbox, dcrBBox, warpRefList, statsCtrl, convergenceMetric, gain, modelWeights, refImage, dcrWeights)
def makeSupplementaryDataGen3(self, butlerQC, inputRefs, outputRefs)
def calculateGain(self, convergenceList, gainList)
def makeSkyInfo(skyMap, tractId, patchId)
Definition: coaddBase.py:279
def calculateModelWeights(self, dcrModels, dcrBBox)
def newModelFromResidual(self, dcrModels, residualGeneratorList, dcrBBox, statsCtrl, gain, modelWeights, refImage, dcrWeights)
def applyModelWeights(self, modelImages, refImage, modelWeights)
def calculateConvergence(self, dcrModels, subExposures, bbox, warpRefList, weightList, statsCtrl)
def run(self, skyInfo, tempExpRefList, imageScalerList, weightList, altMaskList=None, mask=None, supplementaryData=None)
def fillCoadd(self, dcrModels, skyInfo, warpRefList, weightList, calibration=None, coaddInputs=None, mask=None, variance=None)