154 def run(self, expRefList, expDatasetType, imageScalerList=None, refExpDataRef=None, refImageScaler=None):
155 """Match the backgrounds of a list of coadd temp exposures to a reference coadd temp exposure.
157 Choose a refExpDataRef automatically if none supplied.
162 List of data references to science exposures to be background-matched;
163 all exposures must exist.
164 expDatasetType : `str`
165 Dataset type of exposures, e.g. 'goodSeeingCoadd_tempExp'.
166 imageScalerList : `list`, optional
167 List of image scalers (coaddUtils.ImageScaler);
168 if None then the images are not scaled.
169 refExpDataRef : `Unknown`, optional
170 Data reference for the reference exposure.
171 If None, then this task selects the best exposures from expRefList.
172 If not None then must be one of the exposures in expRefList.
173 refImageScaler : `Unknown`, optional
174 Image scaler for reference image;
175 ignored if refExpDataRef is None, else scaling is not performed if None.
179 result : `lsst.pipe.base.Struct`
180 Results as a struct with attributes:
182 ``backgroundInfoList``
183 A `list` of `pipeBase.Struct`, one per exposure in expRefList,
184 each of which contains these fields:
185 - ``isReference``: This is the reference exposure (only one
186 returned Struct will contain True for this
187 value, unless the ref exposure is listed multiple times).
188 - ``backgroundModel``: Differential background model
189 (afw.Math.Background or afw.Math.Approximate).
190 Add this to the science exposure to match the reference exposure.
191 - ``fitRMS``: The RMS of the fit. This is the sqrt(mean(residuals**2)).
192 - ``matchedMSE``: The MSE of the reference and matched images:
193 mean((refImage - matchedSciImage)**2);
194 should be comparable to difference image's mean variance.
195 - ``diffImVar``: The mean variance of the difference image.
196 All fields except isReference will be None if isReference True or the fit failed.
201 Raised if an exposure does not exist on disk.
203 numExp = len(expRefList)
205 raise pipeBase.TaskError(
"No exposures to match")
207 if expDatasetType
is None:
208 raise pipeBase.TaskError(
"Must specify expDatasetType")
210 if imageScalerList
is None:
211 self.log.info(
"imageScalerList is None; no scaling will be performed")
212 imageScalerList = [
None] * numExp
214 if len(expRefList) != len(imageScalerList):
215 raise RuntimeError(
"len(expRefList) = %s != %s = len(imageScalerList)" %
216 (len(expRefList), len(imageScalerList)))
219 if refExpDataRef
is None:
222 expRefList=expRefList,
223 imageScalerList=imageScalerList,
224 expDatasetType=expDatasetType,
226 refExpDataRef = expRefList[refInd]
227 refImageScaler = imageScalerList[refInd]
232 expKeyList = refExpDataRef.butlerSubset.butler.getKeys(expDatasetType)
233 refMatcher =
DataRefMatcher(refExpDataRef.butlerSubset.butler, expDatasetType)
234 refIndSet =
set(refMatcher.matchList(ref0=refExpDataRef, refList=expRefList))
236 if refInd
is not None and refInd
not in refIndSet:
237 raise RuntimeError(
"Internal error: selected reference %s not found in expRefList")
239 refExposure = refExpDataRef.get()
240 if refImageScaler
is not None:
241 refMI = refExposure.getMaskedImage()
242 refImageScaler.scaleMaskedImage(refMI)
244 debugIdKeyList = tuple(
set(expKeyList) -
set([
'tract',
'patch']))
246 self.log.info(
"Matching %d Exposures", numExp)
248 backgroundInfoList = []
249 for ind, (toMatchRef, imageScaler)
in enumerate(zip(expRefList, imageScalerList)):
251 backgroundInfoStruct = pipeBase.Struct(
253 backgroundModel=
None,
259 self.log.info(
"Matching background of %s to %s", toMatchRef.dataId, refExpDataRef.dataId)
261 toMatchExposure = toMatchRef.get()
262 if imageScaler
is not None:
263 toMatchMI = toMatchExposure.getMaskedImage()
264 imageScaler.scaleMaskedImage(toMatchMI)
268 refExposure=refExposure,
269 sciExposure=toMatchExposure,
271 backgroundInfoStruct.isReference =
False
272 except Exception
as e:
273 self.log.warning(
"Failed to fit background %s: %s", toMatchRef.dataId, e)
274 backgroundInfoStruct = pipeBase.Struct(
276 backgroundModel=
None,
282 backgroundInfoList.append(backgroundInfoStruct)
284 return pipeBase.Struct(
285 backgroundInfoList=backgroundInfoList)
289 """Find best exposure to use as the reference exposure.
291 Calculate an appropriate reference exposure by minimizing a cost function that penalizes
292 high variance, high background level, and low coverage. Use the following config parameters:
293 - bestRefWeightCoverage
294 - bestRefWeightVariance
300 List of data references to exposures.
301 Retrieves dataset type specified by expDatasetType.
302 If an exposure is not found, it is skipped with a warning.
303 imageScalerList : `list`
304 List of image scalers (coaddUtils.ImageScaler);
305 must be the same length as expRefList.
306 expDatasetType : `str`
307 Dataset type of exposure: e.g. 'goodSeeingCoadd_tempExp'.
312 Index of best exposure.
317 Raised if none of the exposures in expRefList are found.
319 self.log.info(
"Calculating best reference visit")
321 meanBkgdLevelList = []
324 if len(expRefList) != len(imageScalerList):
325 raise RuntimeError(
"len(expRefList) = %s != %s = len(imageScalerList)" %
326 (len(expRefList), len(imageScalerList)))
328 for expRef, imageScaler
in zip(expRefList, imageScalerList):
329 exposure = expRef.get()
330 maskedImage = exposure.getMaskedImage()
331 if imageScaler
is not None:
333 imageScaler.scaleMaskedImage(maskedImage)
336 varList.append(numpy.nan)
337 meanBkgdLevelList.append(numpy.nan)
338 coverageList.append(numpy.nan)
341 afwMath.MEAN | afwMath.NPOINT | afwMath.VARIANCE, self.
sctrl)
342 meanVar, meanVarErr = statObjIm.getResult(afwMath.VARIANCE)
343 meanBkgdLevel, meanBkgdLevelErr = statObjIm.getResult(afwMath.MEAN)
344 npoints, npointsErr = statObjIm.getResult(afwMath.NPOINT)
345 varList.append(meanVar)
346 meanBkgdLevelList.append(meanBkgdLevel)
347 coverageList.append(npoints)
349 raise pipeBase.TaskError(
350 "None of the candidate %s exist; cannot select best reference exposure" % (expDatasetType,))
353 varArr = numpy.array(varList)/numpy.nanmax(varList)
354 meanBkgdLevelArr = numpy.array(meanBkgdLevelList)/numpy.nanmax(meanBkgdLevelList)
355 coverageArr = numpy.nanmin(coverageList)/numpy.array(coverageList)
357 costFunctionArr = self.config.bestRefWeightVariance * varArr
358 costFunctionArr += self.config.bestRefWeightLevel * meanBkgdLevelArr
359 costFunctionArr += self.config.bestRefWeightCoverage * coverageArr
360 return numpy.nanargmin(costFunctionArr)
363 def matchBackgrounds(self, refExposure, sciExposure):
364 """Match science exposure's background level to that of reference exposure.
366 Process creates a difference image of the reference exposure minus the science exposure, and then
367 generates an afw.math.Background object. It assumes (but does not require/check) that the mask plane
368 already has detections set. If detections have not been set/masked, sources will bias the
369 background estimation.
371 The 'background' of the difference image is smoothed by spline interpolation (by the Background class)
372 or by polynomial interpolation by the Approximate class. This model of difference image
373 is added to the science exposure in memory.
375 Fit diagnostics are also calculated and returned.
379 refExposure : `lsst.afw.image.Exposure`
381 sciExposure : `lsst.afw.image.Exposure`
382 Science exposure; modified by changing the background level
383 to match that of the reference exposure.
387 model : `lsst.pipe.base.Struct`
388 Background model as a struct with attributes:
391 An afw.math.Approximate or an afw.math.Background.
393 RMS of the fit. This is the sqrt(mean(residuals**2)), (`float`).
395 The MSE of the reference and matched images: mean((refImage - matchedSciImage)**2);
396 should be comparable to difference image's mean variance (`float`).
398 The mean variance of the difference image (`float`).
401 refExposure.writeFits(
lsstDebug.Info(__name__).figpath +
'refExposure.fits')
402 sciExposure.writeFits(
lsstDebug.Info(__name__).figpath +
'sciExposure.fits')
405 if self.config.usePolynomial:
406 x, y = sciExposure.getDimensions()
407 shortSideLength =
min(x, y)
408 if shortSideLength < self.config.binSize:
409 raise ValueError(
"%d = config.binSize > shorter dimension = %d" % (self.config.binSize,
411 npoints = shortSideLength // self.config.binSize
412 if shortSideLength % self.config.binSize != 0:
415 if self.config.order > npoints - 1:
416 raise ValueError(
"%d = config.order > npoints - 1 = %d" % (self.config.order, npoints - 1))
419 if (sciExposure.getDimensions() != refExposure.getDimensions()):
420 wSci, hSci = sciExposure.getDimensions()
421 wRef, hRef = refExposure.getDimensions()
423 "Exposures are different dimensions. sci:(%i, %i) vs. ref:(%i, %i)" %
424 (wSci, hSci, wRef, hRef))
426 statsFlag = getattr(afwMath, self.config.gridStatistic)
427 self.
sctrl.setNumSigmaClip(self.config.numSigmaClip)
428 self.
sctrl.setNumIter(self.config.numIter)
430 im = refExposure.getMaskedImage()
431 diffMI = im.Factory(im,
True)
432 diffMI -= sciExposure.getMaskedImage()
434 width = diffMI.getWidth()
435 height = diffMI.getHeight()
436 nx = width // self.config.binSize
437 if width % self.config.binSize != 0:
439 ny = height // self.config.binSize
440 if height % self.config.binSize != 0:
444 bctrl.setUndersampleStyle(self.config.undersampleStyle)
451 if self.config.usePolynomial:
452 order = self.config.order
453 bgX, bgY, bgZ, bgdZ = self.
_gridImage(diffMI, self.config.binSize, statsFlag)
454 minNumberGridPoints =
min(len(
set(bgX)), len(
set(bgY)))
456 raise ValueError(
"No overlap with reference. Nothing to match")
457 elif minNumberGridPoints <= self.config.order:
459 if self.config.undersampleStyle ==
"THROW_EXCEPTION":
460 raise ValueError(
"Image does not cover enough of ref image for order and binsize")
461 elif self.config.undersampleStyle ==
"REDUCE_INTERP_ORDER":
462 self.log.warning(
"Reducing order to %d", (minNumberGridPoints - 1))
463 order = minNumberGridPoints - 1
464 elif self.config.undersampleStyle ==
"INCREASE_NXNYSAMPLE":
465 newBinSize = (minNumberGridPoints*self.config.binSize) // (self.config.order + 1)
466 bctrl.setNxSample(newBinSize)
467 bctrl.setNySample(newBinSize)
469 self.log.warning(
"Decreasing binsize to %d", newBinSize)
472 isUniformImageDiff =
not numpy.any(bgdZ > self.config.gridStdevEpsilon)
473 weightByInverseVariance =
False if isUniformImageDiff
else self.config.approxWeighting
477 if self.config.usePolynomial:
479 order, order, weightByInverseVariance)
480 undersampleStyle = getattr(afwMath, self.config.undersampleStyle)
481 approx = bkgd.getApproximate(actrl, undersampleStyle)
482 bkgdImage = approx.getImage()
484 bkgdImage = bkgd.getImageF(self.config.interpStyle, self.config.undersampleStyle)
485 except Exception
as e:
486 raise RuntimeError(
"Background/Approximation failed to interp image %s: %s" % (
489 sciMI = sciExposure.getMaskedImage()
495 X, Y, Z, dZ = self.
_gridImage(diffMI, self.config.binSize, statsFlag)
496 x0, y0 = diffMI.getXY0()
497 modelValueArr = numpy.empty(len(Z))
498 for i
in range(len(X)):
499 modelValueArr[i] = bkgdImage[int(X[i]-x0), int(Y[i]-y0), afwImage.LOCAL]
500 resids = Z - modelValueArr
501 rms = numpy.sqrt(numpy.mean(resids[~numpy.isnan(resids)]**2))
504 sciExposure.writeFits(
lsstDebug.Info(__name__).figpath +
'sciMatchedExposure.fits')
507 bbox =
geom.Box2D(refExposure.getMaskedImage().getBBox())
509 self.
_debugPlot(X, Y, Z, dZ, bkgdImage, bbox, modelValueArr, resids)
510 except Exception
as e:
511 self.log.warning(
'Debug plot not generated: %s', e)
514 afwMath.MEANCLIP, self.
sctrl).getValue()
516 diffIm = diffMI.getImage()
521 outBkgd = approx
if self.config.usePolynomial
else bkgd
523 return pipeBase.Struct(
524 backgroundModel=outBkgd,
529 def _debugPlot(self, X, Y, Z, dZ, modelImage, bbox, model, resids):
530 """Generate a plot showing the background fit and residuals.
532 It is called when lsstDebug.Info(__name__).savefig = True.
533 Saves the fig to lsstDebug.Info(__name__).figpath.
534 Displays on screen if lsstDebug.Info(__name__).display = True.
538 X : `numpy.ndarray`, (N,)
539 Array of x positions.
540 Y : `numpy.ndarray`, (N,)
541 Array of y positions.
543 Array of the grid values that were interpolated.
544 dZ : `numpy.ndarray`, (len(Z),)
545 Array of the error on the grid values.
546 modelImage : `Unknown`
547 Image of the model of the fit.
548 model : `numpy.ndarray`, (len(Z),)
549 Array of len(Z) containing the grid values predicted by the model.
553 import matplotlib.pyplot
as plt
554 import matplotlib.colors
555 from mpl_toolkits.axes_grid1
import ImageGrid
556 zeroIm = afwImage.MaskedImageF(
geom.Box2I(bbox))
558 x0, y0 = zeroIm.getXY0()
559 dx, dy = zeroIm.getDimensions()
561 self.log.warning(
"No grid. Skipping plot generation.")
563 max, min = numpy.max(Z), numpy.min(Z)
564 norm = matplotlib.colors.normalize(vmax=max, vmin=min)
565 maxdiff = numpy.max(numpy.abs(resids))
566 diffnorm = matplotlib.colors.normalize(vmax=maxdiff, vmin=-maxdiff)
567 rms = numpy.sqrt(numpy.mean(resids**2))
568 fig = plt.figure(1, (8, 6))
569 meanDz = numpy.mean(dZ)
570 grid = ImageGrid(fig, 111, nrows_ncols=(1, 2), axes_pad=0.1,
571 share_all=
True, label_mode=
"L", cbar_mode=
"each",
572 cbar_size=
"7%", cbar_pad=
"2%", cbar_location=
"top")
573 im = grid[0].imshow(zeroIm.getImage().getArray(),
574 extent=(x0, x0+dx, y0+dy, y0), norm=norm,
576 im = grid[0].scatter(X, Y, c=Z, s=15.*meanDz/dZ, edgecolor=
'none', norm=norm,
577 marker=
'o', cmap=
'Spectral')
578 im2 = grid[1].scatter(X, Y, c=resids, edgecolor=
'none', norm=diffnorm,
579 marker=
's', cmap=
'seismic')
580 grid.cbar_axes[0].colorbar(im)
581 grid.cbar_axes[1].colorbar(im2)
582 grid[0].axis([x0, x0+dx, y0+dy, y0])
583 grid[1].axis([x0, x0+dx, y0+dy, y0])
584 grid[0].set_xlabel(
"model and grid")
585 grid[1].set_xlabel(
"residuals. rms = %0.3f"%(rms))
593 """Private method to grid an image for debugging."""
594 width, height = maskedImage.getDimensions()
595 x0, y0 = maskedImage.getXY0()
596 xedges = numpy.arange(0, width, binsize)
597 yedges = numpy.arange(0, height, binsize)
598 xedges = numpy.hstack((xedges, width))
599 yedges = numpy.hstack((yedges, height))
608 for ymin, ymax
in zip(yedges[0:-1], yedges[1:]):
609 for xmin, xmax
in zip(xedges[0:-1], xedges[1:]):
612 subIm = afwImage.MaskedImageF(maskedImage, subBBox, afwImage.PARENT,
False)
614 afwMath.MEAN | afwMath.MEANCLIP | afwMath.MEDIAN
615 | afwMath.NPOINT | afwMath.STDEV,
617 npoints, _ = stats.getResult(afwMath.NPOINT)
619 stdev, _ = stats.getResult(afwMath.STDEV)
620 if stdev < self.config.gridStdevEpsilon:
621 stdev = self.config.gridStdevEpsilon
622 bgX.append(0.5 * (x0 + xmin + x0 + xmax))
623 bgY.append(0.5 * (y0 + ymin + y0 + ymax))
624 bgdZ.append(stdev/numpy.sqrt(npoints))
625 est, _ = stats.getResult(statsFlag)
628 return numpy.array(bgX), numpy.array(bgY), numpy.array(bgZ), numpy.array(bgdZ)