29 import lsst.meas.algorithms
as measAlg
30 import lsst.pex.config
as pexConfig
31 import lsst.pipe.base
as pipeBase
34 from .imageMapReduce
import (ImageMapReduceConfig, ImageMapReduceTask,
37 __all__ = (
"DecorrelateALKernelTask",
"DecorrelateALKernelConfig",
38 "DecorrelateALKernelMapper",
"DecorrelateALKernelMapReduceConfig",
39 "DecorrelateALKernelSpatialConfig",
"DecorrelateALKernelSpatialTask")
43 """Configuration parameters for the DecorrelateALKernelTask
46 ignoreMaskPlanes = pexConfig.ListField(
48 doc=
"""Mask planes to ignore for sigma-clipped statistics""",
49 default=(
"INTRP",
"EDGE",
"DETECTED",
"SAT",
"CR",
"BAD",
"NO_DATA",
"DETECTED_NEGATIVE")
54 """Decorrelate the effect of convolution by Alard-Lupton matching kernel in image difference
59 Pipe-task that removes the neighboring-pixel covariance in an
60 image difference that are added when the template image is
61 convolved with the Alard-Lupton PSF matching kernel.
63 The image differencing pipeline task @link
64 ip.diffim.psfMatch.PsfMatchTask PSFMatchTask@endlink and @link
65 ip.diffim.psfMatch.PsfMatchConfigAL PSFMatchConfigAL@endlink uses
66 the Alard and Lupton (1998) method for matching the PSFs of the
67 template and science exposures prior to subtraction. The
68 Alard-Lupton method identifies a matching kernel, which is then
69 (typically) convolved with the template image to perform PSF
70 matching. This convolution has the effect of adding covariance
71 between neighboring pixels in the template image, which is then
72 added to the image difference by subtraction.
74 The pixel covariance may be corrected by whitening the noise of
75 the image difference. This task performs such a decorrelation by
76 computing a decorrelation kernel (based upon the A&L matching
77 kernel and variances in the template and science images) and
78 convolving the image difference with it. This process is described
79 in detail in [DMTN-021](http://dmtn-021.lsst.io).
81 This task has no standalone example, however it is applied as a
82 subtask of pipe.tasks.imageDifference.ImageDifferenceTask.
84 ConfigClass = DecorrelateALKernelConfig
85 _DefaultName =
"ip_diffim_decorrelateALKernel"
88 """Create the image decorrelation Task
93 arguments to be passed to ``lsst.pipe.base.task.Task.__init__``
95 keyword arguments to be passed to ``lsst.pipe.base.task.Task.__init__``
97 pipeBase.Task.__init__(self, *args, **kwargs)
102 self.
statsControl.setAndMask(afwImage.Mask.getPlaneBitMask(self.config.ignoreMaskPlanes))
106 exposure.getMaskedImage().getMask(),
108 var = statObj.getValue(afwMath.MEANCLIP)
112 def run(self, exposure, templateExposure, subtractedExposure, psfMatchingKernel,
113 preConvKernel=None, xcen=None, ycen=None, svar=None, tvar=None):
114 """Perform decorrelation of an image difference exposure.
116 Decorrelates the diffim due to the convolution of the templateExposure with the
117 A&L PSF matching kernel. Currently can accept a spatially varying matching kernel but in
118 this case it simply uses a static kernel from the center of the exposure. The decorrelation
119 is described in [DMTN-021, Equation 1](http://dmtn-021.lsst.io/#equation-1), where
120 `exposure` is I_1; templateExposure is I_2; `subtractedExposure` is D(k);
121 `psfMatchingKernel` is kappa; and svar and tvar are their respective
122 variances (see below).
126 exposure : `lsst.afw.image.Exposure`
127 The science afwImage.Exposure used for PSF matching
128 templateExposure : `lsst.afw.image.Exposure`
129 The template exposure used for PSF matching
131 the subtracted exposure produced by
132 `ip_diffim.ImagePsfMatchTask.subtractExposures()`
134 An (optionally spatially-varying) PSF matching kernel produced
135 by `ip_diffim.ImagePsfMatchTask.subtractExposures()`
137 if not None, then the `exposure` was pre-convolved with this kernel
138 xcen : `float`, optional
139 X-pixel coordinate to use for computing constant matching kernel to use
140 If `None` (default), then use the center of the image.
141 ycen : `float`, optional
142 Y-pixel coordinate to use for computing constant matching kernel to use
143 If `None` (default), then use the center of the image.
144 svar : `float`, optional
145 image variance for science image
146 If `None` (default) then compute the variance over the entire input science image.
147 tvar : `float`, optional
148 Image variance for template image
149 If `None` (default) then compute the variance over the entire input template image.
154 a `lsst.pipe.base.Struct` containing:
156 - ``correctedExposure`` : the decorrelated diffim
157 - ``correctionKernel`` : the decorrelation correction kernel (which may be ignored)
161 The `subtractedExposure` is NOT updated
163 The returned `correctedExposure` has an updated PSF as well.
165 Here we currently convert a spatially-varying matching kernel into a constant kernel,
166 just by computing it at the center of the image (tickets DM-6243, DM-6244).
168 We are also using a constant accross-the-image measure of sigma (sqrt(variance)) to compute
169 the decorrelation kernel.
171 Still TBD (ticket DM-6580): understand whether the convolution is correctly modifying
172 the variance plane of the new subtractedExposure.
174 spatialKernel = psfMatchingKernel
175 kimg = afwImage.ImageD(spatialKernel.getDimensions())
176 bbox = subtractedExposure.getBBox()
178 xcen = (bbox.getBeginX() + bbox.getEndX()) / 2.
180 ycen = (bbox.getBeginY() + bbox.getEndY()) / 2.
181 self.log.info(
"Using matching kernel computed at (%d, %d)", xcen, ycen)
182 spatialKernel.computeImage(kimg,
True, xcen, ycen)
188 self.log.info(
"Variance (science, template): (%f, %f)", svar, tvar)
193 if np.isnan(svar)
or np.isnan(tvar):
195 if (np.all(np.isnan(exposure.getMaskedImage().getImage().getArray()))
or
196 np.all(np.isnan(templateExposure.getMaskedImage().getImage().getArray()))):
197 self.log.warn(
'Template or science image is entirely NaNs: skipping decorrelation.')
198 outExposure = subtractedExposure.clone()
199 return pipeBase.Struct(correctedExposure=outExposure, correctionKernel=
None)
202 self.log.info(
"Variance (uncorrected diffim): %f", var)
205 if preConvKernel
is not None:
206 self.log.info(
'Using a pre-convolution kernel as part of decorrelation.')
207 kimg2 = afwImage.ImageD(preConvKernel.getDimensions())
208 preConvKernel.computeImage(kimg2,
False)
209 pck = kimg2.getArray()
210 corrKernel = DecorrelateALKernelTask._computeDecorrelationKernel(kimg.getArray(), svar, tvar,
212 correctedExposure, corrKern = DecorrelateALKernelTask._doConvolve(subtractedExposure, corrKernel)
215 psf = subtractedExposure.getPsf().computeKernelImage(
geom.Point2D(xcen, ycen)).getArray()
216 psfc = DecorrelateALKernelTask.computeCorrectedDiffimPsf(corrKernel, psf, svar=svar, tvar=tvar)
217 psfcI = afwImage.ImageD(psfc.shape[0], psfc.shape[1])
218 psfcI.getArray()[:, :] = psfc
220 psfNew = measAlg.KernelPsf(psfcK)
221 correctedExposure.setPsf(psfNew)
224 self.log.info(
"Variance (corrected diffim): %f", var)
226 return pipeBase.Struct(correctedExposure=correctedExposure, correctionKernel=corrKern)
229 def _computeDecorrelationKernel(kappa, svar=0.04, tvar=0.04, preConvKernel=None):
230 """Compute the Lupton decorrelation post-conv. kernel for decorrelating an
231 image difference, based on the PSF-matching kernel.
235 kappa : `numpy.ndarray`
236 A matching kernel 2-d numpy.array derived from Alard & Lupton PSF matching
237 svar : `float`, optional
238 Average variance of science image used for PSF matching
239 tvar : `float`, optional
240 Average variance of template image used for PSF matching
241 preConvKernel If not None, then pre-filtering was applied
242 to science exposure, and this is the pre-convolution kernel.
246 fkernel : `numpy.ndarray`
247 a 2-d numpy.array containing the correction kernel
251 As currently implemented, kappa is a static (single, non-spatially-varying) kernel.
256 kappa = DecorrelateALKernelTask._fixOddKernel(kappa)
257 if preConvKernel
is not None:
258 mk = DecorrelateALKernelTask._fixOddKernel(preConvKernel)
260 if kappa.shape[0] < mk.shape[0]:
261 diff = (mk.shape[0] - kappa.shape[0]) // 2
262 kappa = np.pad(kappa, (diff, diff), mode=
'constant')
263 elif kappa.shape[0] > mk.shape[0]:
264 diff = (kappa.shape[0] - mk.shape[0]) // 2
265 mk = np.pad(mk, (diff, diff), mode=
'constant')
267 kft = np.fft.fft2(kappa)
268 kft2 = np.conj(kft) * kft
269 kft2[np.abs(kft2) < MIN_KERNEL] = MIN_KERNEL
270 denom = svar + tvar * kft2
271 if preConvKernel
is not None:
273 mk2 = np.conj(mk) * mk
274 mk2[np.abs(mk2) < MIN_KERNEL] = MIN_KERNEL
275 denom = svar * mk2 + tvar * kft2
276 denom[np.abs(denom) < MIN_KERNEL] = MIN_KERNEL
277 kft = np.sqrt((svar + tvar) / denom)
278 pck = np.fft.ifft2(kft)
279 pck = np.fft.ifftshift(pck.real)
280 fkernel = DecorrelateALKernelTask._fixEvenKernel(pck)
281 if preConvKernel
is not None:
284 fkernel[fkernel > -np.min(fkernel)] = -np.min(fkernel)
289 fkernel = fkernel[::-1, :]
295 """Compute the (decorrelated) difference image's new PSF.
296 new_psf = psf(k) * sqrt((svar + tvar) / (svar + tvar * kappa_ft(k)**2))
300 kappa : `numpy.ndarray`
301 A matching kernel array derived from Alard & Lupton PSF matching
302 psf : `numpy.ndarray`
303 The uncorrected psf array of the science image (and also of the diffim)
304 svar : `float`, optional
305 Average variance of science image used for PSF matching
306 tvar : `float`, optional
307 Average variance of template image used for PSF matching
311 pcf : `numpy.ndarray`
312 a 2-d numpy.array containing the new PSF
314 def post_conv_psf_ft2(psf, kernel, svar, tvar):
317 if psf.shape[0] < kernel.shape[0]:
318 diff = (kernel.shape[0] - psf.shape[0]) // 2
319 psf = np.pad(psf, (diff, diff), mode=
'constant')
320 elif psf.shape[0] > kernel.shape[0]:
321 diff = (psf.shape[0] - kernel.shape[0]) // 2
322 kernel = np.pad(kernel, (diff, diff), mode=
'constant')
323 psf_ft = np.fft.fft2(psf)
324 kft = np.fft.fft2(kernel)
325 out = psf_ft * np.sqrt((svar + tvar) / (svar + tvar * kft**2))
328 def post_conv_psf(psf, kernel, svar, tvar):
329 kft = post_conv_psf_ft2(psf, kernel, svar, tvar)
330 out = np.fft.ifft2(kft)
333 pcf = post_conv_psf(psf=psf, kernel=kappa, svar=svar, tvar=tvar)
334 pcf = pcf.real / pcf.real.sum()
338 def _fixOddKernel(kernel):
339 """Take a kernel with odd dimensions and make them even for FFT
343 kernel : `numpy.array`
349 a fixed kernel numpy.array. Returns a copy if the dimensions needed to change;
350 otherwise just return the input kernel.
355 if (out.shape[0] % 2) == 1:
356 out = np.pad(out, ((1, 0), (0, 0)), mode=
'constant')
358 if (out.shape[1] % 2) == 1:
359 out = np.pad(out, ((0, 0), (1, 0)), mode=
'constant')
362 out *= (np.mean(kernel) / np.mean(out))
366 def _fixEvenKernel(kernel):
367 """Take a kernel with even dimensions and make them odd, centered correctly.
371 kernel : `numpy.array`
377 a fixed kernel numpy.array
380 maxloc = np.unravel_index(np.argmax(kernel), kernel.shape)
381 out = np.roll(kernel, kernel.shape[0]//2 - maxloc[0], axis=0)
382 out = np.roll(out, out.shape[1]//2 - maxloc[1], axis=1)
384 if (out.shape[0] % 2) == 0:
385 maxloc = np.unravel_index(np.argmax(out), out.shape)
386 if out.shape[0] - maxloc[0] > maxloc[0]:
390 if out.shape[1] - maxloc[1] > maxloc[1]:
397 def _doConvolve(exposure, kernel):
398 """Convolve an Exposure with a decorrelation convolution kernel.
402 exposure : `lsst.afw.image.Exposure`
403 Input exposure to be convolved.
404 kernel : `numpy.array`
405 Input 2-d numpy.array to convolve the image with
409 out : `lsst.afw.image.Exposure`
410 a new Exposure with the convolved pixels and the (possibly
415 We re-center the kernel if necessary and return the possibly re-centered kernel
417 kernelImg = afwImage.ImageD(kernel.shape[0], kernel.shape[1])
418 kernelImg.getArray()[:, :] = kernel
420 maxloc = np.unravel_index(np.argmax(kernel), kernel.shape)
421 kern.setCtrX(maxloc[0])
422 kern.setCtrY(maxloc[1])
423 outExp = exposure.clone()
425 afwMath.convolve(outExp.getMaskedImage(), exposure.getMaskedImage(), kern, convCntrl)
431 """Task to be used as an ImageMapper for performing
432 A&L decorrelation on subimages on a grid across a A&L difference image.
434 This task subclasses DecorrelateALKernelTask in order to implement
435 all of that task's configuration parameters, as well as its `run` method.
438 ConfigClass = DecorrelateALKernelConfig
439 _DefaultName =
'ip_diffim_decorrelateALKernelMapper'
442 DecorrelateALKernelTask.__init__(self, *args, **kwargs)
444 def run(self, subExposure, expandedSubExposure, fullBBox,
445 template, science, alTaskResult=None, psfMatchingKernel=None,
446 preConvKernel=None, **kwargs):
447 """Perform decorrelation operation on `subExposure`, using
448 `expandedSubExposure` to allow for invalid edge pixels arising from
451 This method performs A&L decorrelation on `subExposure` using
452 local measures for image variances and PSF. `subExposure` is a
453 sub-exposure of the non-decorrelated A&L diffim. It also
454 requires the corresponding sub-exposures of the template
455 (`template`) and science (`science`) exposures.
459 subExposure : `lsst.afw.image.Exposure`
460 the sub-exposure of the diffim
461 expandedSubExposure : `lsst.afw.image.Exposure`
462 the expanded sub-exposure upon which to operate
463 fullBBox : `lsst.geom.Box2I`
464 the bounding box of the original exposure
465 template : `lsst.afw.image.Exposure`
466 the corresponding sub-exposure of the template exposure
467 science : `lsst.afw.image.Exposure`
468 the corresponding sub-exposure of the science exposure
469 alTaskResult : `lsst.pipe.base.Struct`
470 the result of A&L image differencing on `science` and
471 `template`, importantly containing the resulting
472 `psfMatchingKernel`. Can be `None`, only if
473 `psfMatchingKernel` is not `None`.
474 psfMatchingKernel : Alternative parameter for passing the
475 A&L `psfMatchingKernel` directly.
476 preConvKernel : If not None, then pre-filtering was applied
477 to science exposure, and this is the pre-convolution
480 additional keyword arguments propagated from
481 `ImageMapReduceTask.run`.
485 A `pipeBase.Struct` containing:
487 - ``subExposure`` : the result of the `subExposure` processing.
488 - ``decorrelationKernel`` : the decorrelation kernel, currently
493 This `run` method accepts parameters identical to those of
494 `ImageMapper.run`, since it is called from the
495 `ImageMapperTask`. See that class for more information.
497 templateExposure = template
498 scienceExposure = science
499 if alTaskResult
is None and psfMatchingKernel
is None:
500 raise RuntimeError(
'Both alTaskResult and psfMatchingKernel cannot be None')
501 psfMatchingKernel = alTaskResult.psfMatchingKernel
if alTaskResult
is not None else psfMatchingKernel
505 subExp2 = scienceExposure.Factory(scienceExposure, expandedSubExposure.getBBox())
506 subExp1 = templateExposure.Factory(templateExposure, expandedSubExposure.getBBox())
509 logLevel = self.log.getLevel()
510 self.log.setLevel(lsst.log.WARN)
511 res = DecorrelateALKernelTask.run(self, subExp2, subExp1, expandedSubExposure,
512 psfMatchingKernel, preConvKernel)
513 self.log.setLevel(logLevel)
515 diffim = res.correctedExposure.Factory(res.correctedExposure, subExposure.getBBox())
516 out = pipeBase.Struct(subExposure=diffim, decorrelationKernel=res.correctionKernel)
521 """Configuration parameters for the ImageMapReduceTask to direct it to use
522 DecorrelateALKernelMapper as its mapper for A&L decorrelation.
524 mapper = pexConfig.ConfigurableField(
525 doc=
'A&L decorrelation task to run on each sub-image',
526 target=DecorrelateALKernelMapper
531 """Configuration parameters for the DecorrelateALKernelSpatialTask.
533 decorrelateConfig = pexConfig.ConfigField(
534 dtype=DecorrelateALKernelConfig,
535 doc=
'DecorrelateALKernel config to use when running on complete exposure (non spatially-varying)',
538 decorrelateMapReduceConfig = pexConfig.ConfigField(
539 dtype=DecorrelateALKernelMapReduceConfig,
540 doc=
'DecorrelateALKernelMapReduce config to use when running on each sub-image (spatially-varying)',
543 ignoreMaskPlanes = pexConfig.ListField(
545 doc=
"""Mask planes to ignore for sigma-clipped statistics""",
546 default=(
"INTRP",
"EDGE",
"DETECTED",
"SAT",
"CR",
"BAD",
"NO_DATA",
"DETECTED_NEGATIVE")
557 """Decorrelate the effect of convolution by Alard-Lupton matching kernel in image difference
562 Pipe-task that removes the neighboring-pixel covariance in an
563 image difference that are added when the template image is
564 convolved with the Alard-Lupton PSF matching kernel.
566 This task is a simple wrapper around @ref DecorrelateALKernelTask,
567 which takes a `spatiallyVarying` parameter in its `run` method. If
568 it is `False`, then it simply calls the `run` method of @ref
569 DecorrelateALKernelTask. If it is True, then it uses the @ref
570 ImageMapReduceTask framework to break the exposures into
571 subExposures on a grid, and performs the `run` method of @ref
572 DecorrelateALKernelTask on each subExposure. This enables it to
573 account for spatially-varying PSFs and noise in the exposures when
574 performing the decorrelation.
576 This task has no standalone example, however it is applied as a
577 subtask of pipe.tasks.imageDifference.ImageDifferenceTask.
578 There is also an example of its use in `tests/testImageDecorrelation.py`.
580 ConfigClass = DecorrelateALKernelSpatialConfig
581 _DefaultName =
"ip_diffim_decorrelateALKernelSpatial"
584 """Create the image decorrelation Task
589 arguments to be passed to
590 `lsst.pipe.base.task.Task.__init__`
592 additional keyword arguments to be passed to
593 `lsst.pipe.base.task.Task.__init__`
595 pipeBase.Task.__init__(self, *args, **kwargs)
600 self.
statsControl.setAndMask(afwImage.Mask.getPlaneBitMask(self.config.ignoreMaskPlanes))
603 """Compute the mean of the variance plane of `exposure`.
606 exposure.getMaskedImage().getMask(),
608 var = statObj.getValue(afwMath.MEANCLIP)
611 def run(self, scienceExposure, templateExposure, subtractedExposure, psfMatchingKernel,
612 spatiallyVarying=True, preConvKernel=None):
613 """Perform decorrelation of an image difference exposure.
615 Decorrelates the diffim due to the convolution of the
616 templateExposure with the A&L psfMatchingKernel. If
617 `spatiallyVarying` is True, it utilizes the spatially varying
618 matching kernel via the `imageMapReduce` framework to perform
619 spatially-varying decorrelation on a grid of subExposures.
623 scienceExposure : `lsst.afw.image.Exposure`
624 the science Exposure used for PSF matching
625 templateExposure : `lsst.afw.image.Exposure`
626 the template Exposure used for PSF matching
627 subtractedExposure : `lsst.afw.image.Exposure`
628 the subtracted Exposure produced by `ip_diffim.ImagePsfMatchTask.subtractExposures()`
630 an (optionally spatially-varying) PSF matching kernel produced
631 by `ip_diffim.ImagePsfMatchTask.subtractExposures()`
632 spatiallyVarying : `bool`
633 if True, perform the spatially-varying operation
634 preConvKernel : `lsst.meas.algorithms.Psf`
635 if not none, the scienceExposure has been pre-filtered with this kernel. (Currently
636 this option is experimental.)
640 results : `lsst.pipe.base.Struct`
641 a structure containing:
643 - ``correctedExposure`` : the decorrelated diffim
646 self.log.info(
'Running A&L decorrelation: spatiallyVarying=%r' % spatiallyVarying)
650 if np.isnan(svar)
or np.isnan(tvar):
652 if (np.all(np.isnan(scienceExposure.getMaskedImage().getImage().getArray()))
or
653 np.all(np.isnan(templateExposure.getMaskedImage().getImage().getArray()))):
654 self.log.warn(
'Template or science image is entirely NaNs: skipping decorrelation.')
663 self.log.info(
"Variance (science, template): (%f, %f)", svar, tvar)
664 self.log.info(
"Variance (uncorrected diffim): %f", var)
665 config = self.config.decorrelateMapReduceConfig
667 results = task.run(subtractedExposure, science=scienceExposure,
668 template=templateExposure, psfMatchingKernel=psfMatchingKernel,
669 preConvKernel=preConvKernel, forceEvenSized=
True)
670 results.correctedExposure = results.exposure
674 return exp.getMaskedImage().getMask()
675 gm(results.correctedExposure)[:, :] = gm(subtractedExposure)
678 self.log.info(
"Variance (corrected diffim): %f", var)
681 config = self.config.decorrelateConfig
683 results = task.run(scienceExposure, templateExposure,
684 subtractedExposure, psfMatchingKernel, preConvKernel=preConvKernel)