lsst.meas.base.noiseReplacer.NoiseReplacer Class Reference

Public Member Functions
def	__init__ (self, config, exposure, footprints, noiseImage=None, exposureId=None, log=None)
def	insertSource (self, id)
def	removeSource (self, id)
def	end (self)
def	getNoiseGenerator (self, exposure, noiseImage, noiseMeanVar, exposureId=None)

Public Attributes
	noiseSource
	noiseOffset
	noiseSeedMultiplier
	noiseGenMean
	noiseGenStd
	removeplanes
	otherbitmask
	heavies
	heavyNoise

Static Public Attributes
	ConfigClass = NoiseReplacerConfig
	exposure = None
	footprints = None
	log = None

Detailed Description

Replace sources with noise during measurement.

Parameters
----------
config : `NoiseReplacerConfig`
    Configuration.
exposure : `lsst.afw.image.Exposure`
    Image in which sources will be replaced by noise. During operation,
    the image will be modified in-place to replace all sources. At the end
    of the measurment procedure, the original sources will be replaced.
footprints : `dict`
    Mapping of ``id`` to a tuple of ``(parent, Footprint)``. When used in
    single-frame measurement, ``id`` is the source ID, but in forced
    photometry this is the reference ID (as that is used to determine
    deblend families).
noiseImage : `lsst.afw.image.ImageF`
    An image used as a predictable noise replacement source. Used during
    testing only.
log : `lsst.log.log.log.Log` or `logging.Logger`, optional
    Logger to use for status messages; no status messages will be recorded
    if `None`.

Notes
-----
When measuring a source (or the children associated with a parent source),
this class is used to replace its neighbors with noise, using the
deblender's definition of the sources as stored in
`~lsst.afw.detection.heavyFootprint.HeavyFootprint`\ s attached to the
`~lsst.afw.table.SourceRecord`\ s.  The algorithm works as follows:

#. All pixels in the source `~lsst.afw.detection.Footprint`\ s are replaced
   with artificially generated noise (in `NoiseReplacer.__init__`).
#. Before each source is measured, we restore the original pixel data by
   inserting that source's
   `~lsst.afw.detection.heavyFootprint.HeavyFootprint` (from the deblender)
   into the image.
#. After measurement, we again replace the source pixels with (the same)
   artificial noise.
#. After measuring all sources, the image is returned to its original
   state.

This is a functional copy of the code in the older
``ReplaceWithNoiseTask``, but with a slightly different API needed for the
new measurement framework; note that it is not an `~lsst.pipe.base.Task`,
as the lifetime of a ``NoiseReplacer`` now corresponds to a single
exposure, not an entire processing run.

When processing the ``footprints`` parameter, this routine should create
`~lsst.afw.detection.heavyFootprint.HeavyFootprint`\ s for any non-Heavy
`~lsst.afw.detection.Footprint`\ s, and replace them in the dictionary. It
should then create a dict of
`~lsst.afw.detection.heavyFootprint.HeavyFootprint`\ s containing noise,
but only for parent objects, then replace all sources with noise. This
should ignore any footprints that lay outside the bounding box of the
exposure, and clip those that lie on the border.

As the code currently stands, the heavy footprint for a deblended object
must be available from the input catalog.  If it is not, it cannot be
reproduced here. In that case, the topmost parent in the objects parent
chain must be used. The heavy footprint for that source is created in
this class from the masked image.

Definition at line 57 of file noiseReplacer.py.

Constructor & Destructor Documentation

__init__()

def lsst.meas.base.noiseReplacer.NoiseReplacer.__init__	(		self,
			config,
			exposure,
			footprints,
			noiseImage = None,
			exposureId = None,
			log = None
	)

Definition at line 135 of file noiseReplacer.py.

    def __init__(self, config, exposure, footprints, noiseImage=None, exposureId=None, log=None):
        noiseMeanVar = None
        self.noiseSource = config.noiseSource
        self.noiseOffset = config.noiseOffset
        self.noiseSeedMultiplier = config.noiseSeedMultiplier
        self.noiseGenMean = None
        self.noiseGenStd = None
        self.log = log
 
        # creates heavies, replaces all footprints with noise
        # We need the source table to be sorted by ID to do the parent lookups
        self.exposure = exposure
        self.footprints = footprints
        mi = exposure.getMaskedImage()
        im = mi.getImage()
        mask = mi.getMask()
        # Add temporary Mask planes for THISDET and OTHERDET
        self.removeplanes = []
        bitmasks = []
        for maskname in ['THISDET', 'OTHERDET']:
            try:
                # does it already exist?
                plane = mask.getMaskPlane(maskname)
                if self.log:
                    self.log.debug('Mask plane "%s" already existed', maskname)
            except Exception:
                # if not, add it; we should delete it when done.
                plane = mask.addMaskPlane(maskname)
                self.removeplanes.append(maskname)
            mask.clearMaskPlane(plane)
            bitmask = mask.getPlaneBitMask(maskname)
            bitmasks.append(bitmask)
            if self.log:
                self.log.debug('Mask plane "%s": plane %i, bitmask %i = 0x%x',
                               maskname, plane, bitmask, bitmask)
        self.thisbitmask, self.otherbitmask = bitmasks
        del bitmasks
        self.heavies = {}
        # Start by creating HeavyFootprints for each source which has no parent
        # and just use them for children which do not already have heavy footprints.
        # If a heavy footprint is available for a child, we will use it. Otherwise,
        # we use the first parent in the parent chain which has a heavy footprint,
        # which with the one level deblender will alway be the topmost parent
        # NOTE: heavy footprints get destroyed by the transform process in forcedPhotCcd.py
        # or forcedPhotCoadd.py so they are never available for forced measurements.
 
        # Create in the dict heavies = {id:heavyfootprint}
        for id, fp in footprints.items():
            if fp[1].isHeavy():
                self.heavies[id] = fp[1]
            elif fp[0] == 0:
                self.heavies[id] = afwDet.makeHeavyFootprint(fp[1], mi)
 
        # ## FIXME: the heavy footprint includes the mask
        # ## and variance planes, which we shouldn't need
        # ## (I don't think we ever want to modify them in
        # ## the input image).  Copying them around is
        # ## wasteful.
 
        # We now create a noise HeavyFootprint for each source with has a heavy footprint.
        # We'll put the noise footprints in a dict heavyNoise = {id:heavyNoiseFootprint}
        self.heavyNoise = {}
        noisegen = self.getNoiseGenerator(exposure, noiseImage, noiseMeanVar, exposureId=exposureId)
        #  The noiseGenMean and Std are used by the unit tests
        self.noiseGenMean = noisegen.mean
        self.noiseGenStd = noisegen.std
        if self.log:
            self.log.debug('Using noise generator: %s', str(noisegen))
        for id in self.heavies:
            fp = footprints[id][1]
            noiseFp = noisegen.getHeavyFootprint(fp)
            self.heavyNoise[id] = noiseFp
            # Also insert the noisy footprint into the image now.
            # Notice that we're just inserting it into "im", ie,
            # the Image, not the MaskedImage.
            noiseFp.insert(im)
            # Also set the OTHERDET bit
            fp.spans.setMask(mask, self.otherbitmask)
 

Member Function Documentation

end()

def lsst.meas.base.noiseReplacer.NoiseReplacer.end

(

self

)

End the NoiseReplacer.

Restores original data to the exposure from the heavies dictionary and
the mask planes to their original state.

Definition at line 274 of file noiseReplacer.py.

    def end(self):
        """End the NoiseReplacer.
 
        Restores original data to the exposure from the heavies dictionary and
        the mask planes to their original state.
        """
        # restores original image, cleans up temporaries
        # (ie, replace all the top-level pixels)
        mi = self.exposure.getMaskedImage()
        im = mi.getImage()
        mask = mi.getMask()
        for id in self.footprints.keys():
            if self.footprints[id][0] != 0:
                continue
            self.heavies[id].insert(im)
        for maskname in self.removeplanes:
            mask.removeAndClearMaskPlane(maskname, True)
 
        del self.removeplanes
        del self.thisbitmask
        del self.otherbitmask
        del self.heavies
        del self.heavyNoise
 

getNoiseGenerator()

def lsst.meas.base.noiseReplacer.NoiseReplacer.getNoiseGenerator	(		self,
			exposure,
			noiseImage,
			noiseMeanVar,
			exposureId = None
	)

Return a generator of artificial noise.

Returns
-------
noiseGenerator : `lsst.afw.image.noiseReplacer.NoiseGenerator`

Definition at line 298 of file noiseReplacer.py.

    def getNoiseGenerator(self, exposure, noiseImage, noiseMeanVar, exposureId=None):
        """Return a generator of artificial noise.
 
        Returns
        -------
        noiseGenerator : `lsst.afw.image.noiseReplacer.NoiseGenerator`
        """
        if noiseImage is not None:
            return ImageNoiseGenerator(noiseImage)
        rand = None
        if self.noiseSeedMultiplier:
            # default plugin, our seed
            if exposureId is not None and exposureId != 0:
                seed = exposureId*self.noiseSeedMultiplier
            else:
                seed = self.noiseSeedMultiplier
            rand = afwMath.Random(afwMath.Random.MT19937, seed)
        if noiseMeanVar is not None:
            try:
                # Assume noiseMeanVar is an iterable of floats
                noiseMean, noiseVar = noiseMeanVar
                noiseMean = float(noiseMean)
                noiseVar = float(noiseVar)
                noiseStd = math.sqrt(noiseVar)
                if self.log:
                    self.log.debug('Using passed-in noise mean = %g, variance = %g -> stdev %g',
                                   noiseMean, noiseVar, noiseStd)
                return FixedGaussianNoiseGenerator(noiseMean, noiseStd, rand=rand)
            except Exception:
                if self.log:
                    self.log.debug('Failed to cast passed-in noiseMeanVar to floats: %s',
                                   str(noiseMeanVar))
        offset = self.noiseOffset
        noiseSource = self.noiseSource
 
        if noiseSource == 'meta':
            # check the exposure metadata
            meta = exposure.getMetadata()
            # this key name correspond to SubtractBackgroundTask() in meas_algorithms
            try:
                bgMean = meta.getAsDouble('BGMEAN')
                # We would have to adjust for GAIN if ip_isr didn't make it 1.0
                noiseStd = math.sqrt(bgMean)
                if self.log:
                    self.log.debug('Using noise variance = (BGMEAN = %g) from exposure metadata',
                                   bgMean)
                return FixedGaussianNoiseGenerator(offset, noiseStd, rand=rand)
            except Exception:
                if self.log:
                    self.log.debug('Failed to get BGMEAN from exposure metadata')
 
        if noiseSource == 'variance':
            if self.log:
                self.log.debug('Will draw noise according to the variance plane.')
            var = exposure.getMaskedImage().getVariance()
            return VariancePlaneNoiseGenerator(var, mean=offset, rand=rand)
 
        # Compute an image-wide clipped variance.
        im = exposure.getMaskedImage().getImage()
        s = afwMath.makeStatistics(im, afwMath.MEANCLIP | afwMath.STDEVCLIP)
        noiseMean = s.getValue(afwMath.MEANCLIP)
        noiseStd = s.getValue(afwMath.STDEVCLIP)
        if self.log:
            self.log.debug("Measured from image: clipped mean = %g, stdev = %g",
                           noiseMean, noiseStd)
        return FixedGaussianNoiseGenerator(noiseMean + offset, noiseStd, rand=rand)
 
 

insertSource()

def lsst.meas.base.noiseReplacer.NoiseReplacer.insertSource	(		self,
			id
	)

Insert the heavy footprint of a given source into the exposure.

Parameters
----------
id : `int`
    ID of the source to insert from original dictionary of footprints.

Notes
-----
Also adjusts the mask plane to show the source of this footprint.

Definition at line 214 of file noiseReplacer.py.

    def insertSource(self, id):
        """Insert the heavy footprint of a given source into the exposure.
 
        Parameters
        ----------
        id : `int`
            ID of the source to insert from original dictionary of footprints.
 
        Notes
        -----
        Also adjusts the mask plane to show the source of this footprint.
        """
        # Copy this source's pixels into the image
        mi = self.exposure.getMaskedImage()
        im = mi.getImage()
        mask = mi.getMask()
        # usedid can point either to this source, or to the first parent in the
        # parent chain which has a heavy footprint (or to the topmost parent,
        # which always has one)
        usedid = id
        while self.footprints[usedid][0] != 0 and usedid not in self.heavies:
            usedid = self.footprints[usedid][0]
        fp = self.heavies[usedid]
        fp.insert(im)
        fp.spans.setMask(mask, self.thisbitmask)
        fp.spans.clearMask(mask, self.otherbitmask)
 

removeSource()

def lsst.meas.base.noiseReplacer.NoiseReplacer.removeSource	(		self,
			id
	)

Replace the heavy footprint of a given source with noise.

The same artificial noise is used as in the original replacement.

Parameters
----------
id : `int`
    ID of the source to replace from original dictionary of footprints.

Notes
-----
Also restores the mask plane.

Definition at line 241 of file noiseReplacer.py.

    def removeSource(self, id):
        """Replace the heavy footprint of a given source with noise.
 
        The same artificial noise is used as in the original replacement.
 
        Parameters
        ----------
        id : `int`
            ID of the source to replace from original dictionary of footprints.
 
        Notes
        -----
        Also restores the mask plane.
        """
        # remove a single source
        # (Replace this source's pixels by noise again.)
        # Do this by finding the source's top-level ancestor
        mi = self.exposure.getMaskedImage()
        im = mi.getImage()
        mask = mi.getMask()
 
        # use the same algorithm as in remove Source to find the heavy noise footprint
        # which will undo what insertSource(id) does
        usedid = id
        while self.footprints[usedid][0] != 0 and usedid not in self.heavies:
            usedid = self.footprints[usedid][0]
        # Re-insert the noise pixels
        fp = self.heavyNoise[usedid]
        fp.insert(im)
        # Clear the THISDET mask plane.
        fp.spans.clearMask(mask, self.thisbitmask)
        fp.spans.setMask(mask, self.otherbitmask)
 

Member Data Documentation

ConfigClass

lsst.meas.base.noiseReplacer.NoiseReplacer.ConfigClass = NoiseReplacerConfig

static

Definition at line 121 of file noiseReplacer.py.

exposure

lsst.meas.base.noiseReplacer.NoiseReplacer.exposure = None

static

Definition at line 123 of file noiseReplacer.py.

footprints

lsst.meas.base.noiseReplacer.NoiseReplacer.footprints = None

static

Definition at line 127 of file noiseReplacer.py.

heavies

lsst.meas.base.noiseReplacer.NoiseReplacer.heavies

Definition at line 172 of file noiseReplacer.py.

heavyNoise

lsst.meas.base.noiseReplacer.NoiseReplacer.heavyNoise

Definition at line 196 of file noiseReplacer.py.

log

lsst.meas.base.noiseReplacer.NoiseReplacer.log = None

static

Definition at line 131 of file noiseReplacer.py.

noiseGenMean

lsst.meas.base.noiseReplacer.NoiseReplacer.noiseGenMean

Definition at line 140 of file noiseReplacer.py.

noiseGenStd

lsst.meas.base.noiseReplacer.NoiseReplacer.noiseGenStd

Definition at line 141 of file noiseReplacer.py.

noiseOffset

lsst.meas.base.noiseReplacer.NoiseReplacer.noiseOffset

Definition at line 138 of file noiseReplacer.py.

noiseSeedMultiplier

lsst.meas.base.noiseReplacer.NoiseReplacer.noiseSeedMultiplier

Definition at line 139 of file noiseReplacer.py.

noiseSource

lsst.meas.base.noiseReplacer.NoiseReplacer.noiseSource

Definition at line 137 of file noiseReplacer.py.

otherbitmask

lsst.meas.base.noiseReplacer.NoiseReplacer.otherbitmask

Definition at line 170 of file noiseReplacer.py.

removeplanes

lsst.meas.base.noiseReplacer.NoiseReplacer.removeplanes

Definition at line 152 of file noiseReplacer.py.

---

The documentation for this class was generated from the following file:

• /j/snowflake/release/lsstsw/stack/lsst-scipipe-0.7.0/Linux64/meas_base/22.0.1-19-g93a5c75+d23f2fb6d8/python/lsst/meas/base/noiseReplacer.py