def _correctBadPix(self,badmask):
for i in np.arange(self.stack.shape[2]):
plane = self.stack[:,:,i]
maskedImage = np.ma.array(plane, mask=badmask.mask)
NANMask = maskedImage.filled(np.NaN)
self.stack[:,:,i] = inpaint.replace_nans(NANMask, 5, 0.5, 1, 'idw')
plane = self.ustack[:,:,i]
maskedImage = np.ma.array(plane, mask=badmask.mask)
NANMask = maskedImage.filled(np.NaN)
self.ustack[:,:,i] = inpaint.replace_nans(NANMask, 5, 0.5, 1, 'idw')
def inpaint_array(inputArray, mask, **kwargs):
maskedImg = np.ma.array(inputArray, mask=mask)
NANMask = maskedImg.filled(np.NaN)
badArrays, num_badArrays = ndimage.label(mask)
# data_slices = ndimage.find_objects(badArrays)
filled = inpaint.replace_nans(NANMask, **kwargs)
return filled
def inpaint(self):
""" Replace masked-out elements in an array using an iterative image inpainting algorithm. """
import inpaint
filled = inpaint.replace_nans(
np.ma.filled(self.raster_data, np.NAN).astype(np.float32), 3, 0.01,
2)
self.raster_data = np.ma.masked_invalid(filled)
def callInpaint(img, mask, outputFilename, i):
maskedImg = np.ma.array(img, mask = mask)
NANMask = maskedImg.filled(np.NaN)
filled = inpaint.replace_nans(NANMask)
#outputFilename = utils.createOutputFilename(sdssFilename, dataDir)
hdu = pyfits.PrimaryHDU(filled)
if os.path.exists(outputFilename):
print outputFilename
outputFilename = outputFilename+"B"
hdu.writeto(outputFilename)
else:
hdu.writeto(outputFilename)
def inpaint_array(inputArray, mask):
# maskedImg = np.ma.array(inputArray, mask=mask)
# NANMask = maskedImg.filled(np.NaN)
# badArrays, num_badArrays = ndimage.label(mask)
# data_slices = ndimage.find_objects(badArrays)
filled = replace_nans(
inputArray * mask,
max_iter=20,
tol=0.05,
kernel_radius=5,
kernel_sigma=2,
method='idw')
return filled
def runInpainting(maskFile, listFile, dataDir, ID, log):
maskFilename = utils.getMask(maskFile, ID)
sdssFilename = GalaxyParameters.getSDSSUrl(listFile, dataDir, ID)
outputFilename = utils.createOutputFilename(sdssFilename)
sdssImage = pyfits.open(sdssFilename)
imageData = sdssImage[0].data
imageData = imageData - 1000 #soft bias subtraction, comment out if needed
head = sdssImage[0].header
maskFile = pyfits.open(maskFilename)
mask = maskFile[0].data
maskedImg = np.ma.array(imageData, mask = mask)
NANMask = maskedImg.filled(np.NaN)
filled = inpaint.replace_nans(NANMask, 15, 0.1, 2, 'idw')
hdu = pyfits.PrimaryHDU(filled, header = head)
try:
hdu.writeto(dataDir+'/filled/'+outputFilename)
except IOError:
errmsg = 'file already exists!', dataDir+'/filled/'+outputFilename
print errmsg
log.append(errmsg)
return log
def inpaint(self):
""" Replace masked-out elements in an array using an iterative image inpainting algorithm. """
import inpaint
filled = inpaint.replace_nans(np.ma.filled(self.raster_data, np.NAN).astype(np.float32), 3, 0.01, 2)
self.raster_data = np.ma.masked_invalid(filled)
def saveDataClearMask(edfimagepath,
newpath,
stem,
first,
last,
medianename,
mask,
ndigits=4,
extension='edf',
doinpaint=False):
"""
Save new EDF files with the median and mask removed
edfimagepath: Path to the EDF images. The median image is assumed to be in the same directory
newpath: Path to save the new EDF images.
stem: Stem for EDF images.
first: First image number
last: Last image number
medianename: Name of median EDF file (in full, with extension)
mask: mask data
ndigits: Number of digits for EDF file numbering.
extension: EDF file extension.
doinpaint: if set to true, fills diamond mask with inpainting. If not set, diamond mask is filled with median value
"""
if ((not (os.path.isdir(newpath))) or (not (os.path.exists(newpath)))):
print("ERROR! %s is not a directory or does not exist.\nAborting." %
newpath)
return
if (os.path.samefile(edfimagepath, newpath)):
print(
"ERROR!\nImages are read from %s.\nNew EDF should be saved in %s.\nThis will destroy the original data.\nAborting"
% (edfimagepath, newpath))
return
print("Reading median image")
# Read median image
imagename = os.path.join(edfimagepath, medianename)
medianeIm = fabio.edfimage.edfimage()
medianeIm.read(imagename)
medianeData = medianeIm.getData().astype('float32')
if (doinpaint):
print("Filling mask with inpainting")
else:
print("Filling mask with median value")
# Loop on images and test median substraction
for i in range(first, last + 1):
# Read image data
format = "%s%0" + str(ndigits) + "d." + extension
image = format % (stem, i)
imagename = os.path.join(edfimagepath, image)
print("Reading and processing " + imagename)
im = fabio.edfimage.edfimage()
im.read(imagename)
data = im.getData().astype('float32')
header = im.getHeader()
# Removing median image
data = data - medianeData
# Removing anything below 0
data = data.clip(min=0)
meanI = data.mean()
medianI = numpy.median(data)
maxI = data.max()
minI = data.min()
xsize = im.getDim1()
ysize = im.getDim2()
# Preparing mask
thismask = mask[i - first]
thismask = thismask.astype(
numpy.float32
) # New versions of python do not like resizing with integer...
maskscaled = scipy.misc.imresize(thismask, (xsize, ysize),
interp='nearest',
mode='F')
# Creating data under mask using linear interpolation or inpainting
# Need to create a list of points for which we have data
# Actually, gave up, fill with median value!
idx = (maskscaled > 0)
if (doinpaint):
# Creating data under mask using inpainting
# We rescale the image and inpaint on a smaller version. Before reducing, remove extreme intensity values (it works better)
datacopy = data.copy()
datacopy = datacopy.clip(0., 10. * meanI)
datareduced = scipy.misc.imresize(
datacopy, (thismask.shape[0], thismask.shape[1]),
interp='nearest',
mode='F')
# Setting mask data as NaN and call for inpainting. Parameters have been set from trial and error
idx2 = (thismask > 0)
datareduced[idx2] = numpy.NaN
result0 = inpaint.replace_nans(datareduced,
max_iter=20,
tol=1.,
kernel_radius=2,
kernel_sigma=5,
method='idw')
# Rescaling inpainted image and set new values at mask positions
result0 = scipy.misc.imresize(result0, (xsize, ysize),
interp='nearest',
mode='F')
data[idx] = result0[idx]
else:
# Fill maslwith median value!
data[idx] = medianI
# Save new data
newname = os.path.join(newpath, image)
print("Saving new EDF with median and mask removed in " + newname)
im = fabio.edfimage.edfimage()
im.read(imagename)
im.setData(data.astype('uint32'))
im.setHeader(header)
im.save(newname)
NANMask = maskedImg.filled(np.NaN)
badArrays, num_badArrays = sp.ndimage.label(mask)
print num_badArrays
data_slices = sp.ndimage.find_objects(badArrays)
#data_slices = np.ma.extras.notmasked_contiguous(maskedImg)
import inpaint
#def replace_nans(array, max_iter, tol,kernel_size=1,method='localmean'):
#data = np.array([[11,22.0,353],[1,np.NaN,343],[11,0,343]])
filled = inpaint.replace_nans(NANMask, 5, 0.5, 2, 'idw')
hdu = pyfits.PrimaryHDU(filled, header = head)
hdu.writeto('IC3376R_masked.fits')
#hdu = pyfits.PrimaryHDU(NANMask)
#hdu.writeto('NanMask.fits')
#invdisttree = idw.Invdisttree(maskedImg, maskedImg)
#print data_slices
#for slice in data_slices:
# dy, dx = slice
# coords = float(dy.start), float(dy.stop), float(dx.start), float(dx.stop)
# print 'slice', coords
# submask = cutoutMask[(dy.start):(dy.stop),(dx.start):(dx.stop)]
# subarray = cutout[(dy.start):(dy.stop+1),(dx.start):(dx.stop+1)]
