def lacos(_input0, output='clean.fits', outmask='mask.fits', gain=1.3, readn=9, xorder=9, yorder=9, sigclip=4.5, sigfrac=0.5, objlim=1, verbose=True, interactive=False): import floyds from floyds.util import delete from pyraf import iraf import numpy as np oldoutput, galaxy, skymod, med5 = 'oldoutput.fits', 'galaxy.fits', 'skymod.fits', 'med5.fits' blk, lapla, med3, med7, sub5, sigima, finalsel = 'blk.fits', 'lapla.fits', 'med3.fits', 'med7.fits', 'sub5.fits',\ 'sigima.fits', 'finalsel.fits' deriv2, noise, sigmap, firstsel, starreject = 'deriv2.fits', 'noise.fits', 'sigmap.fits', 'firstsel.fits',\ 'starreject.fits' inputmask = 'inputmask.fits' # set some parameters in standard IRAF tasks iraf.convolve.bilinear = 'no' iraf.convolve.radsym = 'no' # create Laplacian kernel # laplkernel = np.array([[0.0, -1.0, 0.0], [-1.0, 4.0, -1.0], [0.0, -1.0, 0.0]]) f = open('_kernel', 'w') f.write('0 -1 0;\n-1 4 -1;\n0 -1 0') f.close() # create growth kernel f = open('_gkernel', 'w') f.write('1 1 1;\n1 1 1;\n1 1 1') f.close() gkernel = np.array([[1.0, 1.0, 1.0], [1.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) delete(galaxy) delete(skymod) delete(oldoutput) if not output: output = _input0 else: delete(output) iraf.imcopy(_input0, output, verbose='no') delete('_xxx.fits,_yyy.fits') iraf.imcopy(_input0 + '[*,10:80]', '_xxx.fits', verbose='no') _input = '_xxx.fits' arrayinput, headerinput = floyds.cosmics.fromfits(_input, verbose=False) floyds.cosmics.tofits(outmask, np.float32(arrayinput - arrayinput), headerinput, verbose=False) # subtract object spectra if desired # iraf.fit1d(_input,galaxy,"fit",axis=2,order=5,func="leg",low=4., # high=4.,nav=1,inter='yes',sample="*",niter=3,grow=0,cursor="") # iraf.imarith(_input,"-",galaxy,oldoutput) # iraf.display(oldoutput,1,fill='yes') ####### #Subtract sky lines # iraf.fit1d(oldoutput,skymod,"fit",axis=2,order=5,func="leg",low=4.,high=4., # inter='no',sample="*",nav=1,niter=3,grow=0,cursor="") # iraf.imarith(oldoutput,"-",skymod,oldoutput) # iraf.display(oldoutput,2,fill='yes') ##### iraf.imcopy(_input, oldoutput) arrayoldoutput, headeroldoutput = floyds.cosmics.fromfits(oldoutput, verbose=False) # add object spectra to sky model # iraf.imarith(skymod,"+",galaxy,skymod) delete(med5) # add median of residuals to sky model iraf.median(oldoutput, med5, 5, 5, zlor='INDEF', zhir='INDEF', verbose='no') # m5 = ndimage.filters.median_filter(_inputarray, size=5, mode='mirror') # iraf.imarith(skymod,"+",med5,med5) # take second-order derivative (Laplacian) of input image # kernel is convolved with subsampled image, in order to remove negative # pattern around high pixels delete(blk) delete(lapla) iraf.blkrep(oldoutput, blk, 2, 2) iraf.convolve(blk, lapla, '_kernel') iraf.imreplace(lapla, 0, upper=0, lower='INDEF') delete(deriv2) delete(noise) iraf.blkavg(lapla, deriv2, 2, 2, option="average") # create noise model iraf.imutil.imexpr(expr='sqrt(a*' + str(gain) + '+' + str(readn) + '**2)/' + str(gain), a=med5, output=noise, verbose='no') iraf.imreplace(med5, 0.00001, upper=0, lower='INDEF') # divide Laplacian by noise model delete(sigmap) iraf.imutil.imexpr(expr='(a/b)/2', a=deriv2, b=noise, output=sigmap, verbose='no') # removal of large structure (bright, extended objects) delete(med5) iraf.median(sigmap, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imarith(sigmap, "-", med5, sigmap) # find all candidate cosmic rays # this selection includes sharp features such as stars and HII regions arraysigmap, headersigmap = floyds.cosmics.fromfits(sigmap, verbose=False) arrayf = np.where(arraysigmap < sigclip, 0, arraysigmap) arrayf = np.where(arrayf > 0.1, 1, arrayf) floyds.cosmics.tofits(firstsel, np.float32(arrayf), headersigmap, verbose=False) # compare candidate CRs to median filtered image # this step rejects bright, compact sources from the initial CR list # subtract background and smooth component of objects delete(med3) iraf.median(oldoutput, med3, 3, 3, zlo='INDEF', zhi='INDEF', verbose='no') delete(med7) delete('_' + med3) iraf.median(med3, med7, 7, 7, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imutil.imexpr(expr='(a-b)/c', a=med3, b=med7, c=noise, output='_' + med3, verbose='no') iraf.imreplace('_' + med3, 0.01, upper=0.01, lower='INDEF') # compare CR flux to object flux delete(starreject) iraf.imutil.imexpr(expr='a+b+c', a=firstsel, b=sigmap, c="_" + med3, output=starreject, verbose='no') # discard if CR flux <= objlim * object flux iraf.imreplace(starreject, 0, upper=objlim, lower='INDEF') iraf.imreplace(starreject, 1, lower=0.5, upper='INDEF') iraf.imarith(firstsel, "*", starreject, firstsel) # grow CRs by one pixel and check in original sigma map arrayfirst, headerfirst = floyds.cosmics.fromfits(firstsel, verbose=False) arraygfirst = floyds.cosmics.my_convolve_with_FFT2(arrayfirst, gkernel) arraygfirst = np.where(arraygfirst > 0.5, 1, arraygfirst) arraygfirst = arraygfirst * arraysigmap arraygfirst = np.where(arraygfirst < sigclip, 0, arraygfirst) arraygfirst = np.where(arraygfirst > 0.1, 1, arraygfirst) # grow CRs by one pixel and lower detection limit sigcliplow = sigfrac * sigclip # Finding neighbouring pixels affected by cosmic rays arrayfinal = floyds.cosmics.my_convolve_with_FFT2(arraygfirst, gkernel) arrayfinal = np.where(arrayfinal > 0.5, 1, arrayfinal) arrayfinal = arrayfinal * arraysigmap arrayfinal = np.where(arrayfinal < sigcliplow, 0, arrayfinal) arrayfinal = np.where(arrayfinal > 0.1, 1, arrayfinal) # determine number of CRs found in this iteration arraygfirst = (1 - (arrayfinal - arrayfinal)) * arrayfinal npix = [str(int(np.size(np.where(arraygfirst > 0.5)) / 2.))] # create cleaned output image; use 3x3 median with CRs excluded arrayoutmask = np.where(arrayfinal > 1, 1, arrayfinal) floyds.cosmics.tofits(outmask, np.float32(arrayoutmask), headerfirst, verbose=False) delete(inputmask) arrayinputmask = (1 - (10000 * arrayoutmask)) * arrayoldoutput floyds.cosmics.tofits(inputmask, np.float32(arrayinputmask), headerfirst, verbose=False) delete(med5) iraf.median(inputmask, med5, 5, 5, zloreject=-9999, zhi='INDEF', verbose='no') iraf.imarith(outmask, "*", med5, med5) delete('_yyy.fits') iraf.imutil.imexpr(expr='(1-a)*b+c', a=outmask, b=oldoutput, c=med5, output='_yyy.fits', verbose='no') # add sky and object spectra back in #iraf.imarith('_yyy.fits',"+",skymod,'_yyy.fits') # cleanup and get ready for next iteration if npix == 0: stop = yes # delete temp files iraf.imcopy('_yyy.fits', output + '[*,10:80]', verbose='no') delete(blk + "," + lapla + "," + deriv2 + "," + med5) delete(med3 + "," + med7 + "," + noise + "," + sigmap) delete(firstsel + "," + starreject) delete(finalsel + "," + inputmask) delete(oldoutput + "," + skymod + "," + galaxy) delete("_" + med3 + ",_" + sigmap) delete('_kernel' + "," + '_gkernel') delete(outmask) delete('_xxx.fits,_yyy.fits')
def lacos_im(_input, _output='clean.fits', outmask='mask.fits', gain=1.3, readn=9, xorder=9, yorder=9, sigclip=4.5, sigfrac=0.5, objlim=1, skyval=0, niter=2, verbose=True, interactive=False): import floyds from floyds.util import delete import sys, re, os, string from pyraf import iraf import numpy as np iraf.convolve.bilinear = 'no' iraf.convolve.radsym = 'no' # make temporary files oldoutput, galaxy, skymod, med5 = 'oldoutput.fits', 'galaxy.fits', 'skymod.fits', 'med5.fits' blk, lapla, med3, med7, sub5, sigima, finalsel = 'blk.fits', 'lapla.fits', 'med3.fits', 'med7.fits', 'sub5.fits',\ 'sigima.fits', 'finalsel.fits' deriv2, noise, sigmap, firstsel, starreject = 'deriv2.fits', 'noise.fits', 'sigmap.fits', 'firstsel.fits',\ 'starreject.fits' inputmask, gfirstsel = 'inputmask.fits', 'gfirstsel.fits' f = open('_kernel', 'w') f.write('0 -1 0;\n-1 4 -1;\n0 -1 0') f.close() # create growth kernel f = open('_gkernel', 'w') f.write('1 1 1;\n1 1 1;\n1 1 1') f.close() gkernel = np.array([[1.0, 1.0, 1.0], [1.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) # initialize loop usegain = gain i = 1 stop = 'no' previous = 0 if not _output: _output = _input arrayinput, headerinput = floyds.cosmics.fromfits(_input, verbose=False) floyds.cosmics.tofits(outmask, np.float32(arrayinput - arrayinput), headerinput, verbose=False) delete(oldoutput) if skyval > 0: arrayoldoutput = arrayinput + skyval else: arrayoldoutput = arrayinput floyds.cosmics.tofits(oldoutput, np.float32(arrayoldoutput), headerinput, verbose=False) # start iterations while stop == 'no': # take second-order derivative (Laplacian) of input image # kernel is convolved with subsampled image, in order to remove negative # pattern around high pixels delete(blk) delete(lapla) delete(deriv2) iraf.blkrep(oldoutput, blk, 2, 2) iraf.convolve(blk, lapla, '_kernel') iraf.imreplace(lapla, 0, upper=0, lower='INDEF', radius=0) iraf.blkavg(lapla, deriv2, 2, 2, option="average") delete(med5) # create model of background flux - 5x5 box should exclude all CRs iraf.median(oldoutput, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imreplace(med5, 0.0001, upper=0, lower='INDEF', radius=0) # create noise model delete(noise) iraf.imutil.imexpr(expr='sqrt(a*' + str(usegain) + '+' + str(readn) + '**2)/' + str(usegain), a=med5, output=noise, verbose='no') # divide Laplacian by noise model delete(sigmap) iraf.imarith(deriv2, "/", noise, sigmap) # Laplacian of blkreplicated image counts edges twice: iraf.imarith(sigmap, "/", 2., sigmap) # removal of large structure (bright, extended objects) delete(med5) iraf.median(sigmap, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') arraysigmap, headersigmap = floyds.cosmics.fromfits(sigmap, verbose=False) arraymed5, headermed5 = floyds.cosmics.fromfits(med5, verbose=False) arraysigmap = arraysigmap - arraymed5 iraf.imarith(sigmap, "-", med5, sigmap) # find all candidate cosmic rays # this selection includes sharp features such as stars and HII regions delete(firstsel) iraf.imcopy(sigmap, firstsel, verbose='no') iraf.imreplace(firstsel, 0, upper=sigclip, lower='INDEF', radius=0) iraf.imreplace(firstsel, 1, lower=0.1, upper='INDEF', radius=0) # arraygfirst=arraysigmap # arraygfirst = np.where(arraygfirst<sigclip,0,arraygfirst) # arraygfirst = np.where(arraygfirst>0.1,1,arraygfirst) # compare candidate CRs to median filtered image # this step rejects bright, compact sources from the initial CR list # subtract background and smooth component of objects delete(med3) delete(med7) iraf.median(oldoutput, med3, 3, 3, zlo='INDEF', zhi='INDEF', verbose='no') iraf.median(med3, med7, 7, 7, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imarith(med3, "-", med7, med3) iraf.imarith(med3, "/", noise, med3) iraf.imreplace(med3, 0.01, upper=0.01, lower='INDEF', radius=0) # compare CR flux to object flux delete(starreject) iraf.imutil.imexpr(expr="(a*b)/c", a=firstsel, b=sigmap, c=med3, output=starreject, verbose='no') # discard if CR flux <= objlim * object flux iraf.imreplace(starreject, 0, upper=objlim, lower='INDEF', radius=0) iraf.imreplace(starreject, 1, lower=0.5, upper='INDEF', radius=0) iraf.imarith(firstsel, "*", starreject, firstsel) # grow CRs by one pixel and check in original sigma map delete(gfirstsel) iraf.convolve(firstsel, gfirstsel, '_gkernel') iraf.imreplace(gfirstsel, 1, lower=0.5, upper='INDEF', radius=0) iraf.imarith(gfirstsel, "*", sigmap, gfirstsel) iraf.imreplace(gfirstsel, 0, upper=sigclip, lower='INDEF', radius=0) iraf.imreplace(gfirstsel, 1, lower=0.1, upper='INDEF', radius=0) # grow CRs by one pixel and lower detection limit sigcliplow = sigfrac * sigclip delete(finalsel) iraf.convolve(gfirstsel, finalsel, '_gkernel') iraf.imreplace(finalsel, 1, lower=0.5, upper='INDEF', radius=0) iraf.imarith(finalsel, "*", sigmap, finalsel) iraf.imreplace(finalsel, 0, upper=sigcliplow, lower='INDEF', radius=0) iraf.imreplace(finalsel, 1, lower=0.1, upper='INDEF', radius=0) # determine number of CRs found in this iteration delete(gfirstsel) iraf.imutil.imexpr(expr="(1-b)*a", a=outmask, b=finalsel, output=gfirstsel, verbose='no') npix = iraf.imstat(gfirstsel, fields="npix", lower=0.5, upper='INDEF', Stdout=1) # create cleaned output image; use 3x3 median with CRs excluded delete(med5) iraf.imarith(outmask, "+", finalsel, outmask) iraf.imreplace(outmask, 1, lower=1, upper='INDEF', radius=0) delete(inputmask) iraf.imutil.imexpr(expr="(1-10000*a)", a=outmask, output=inputmask, verbose='no') iraf.imarith(oldoutput, "*", inputmask, inputmask) delete(med5) iraf.median(inputmask, med5, 5, 5, zloreject=-9999, zhi='INDEF', verbose='no') iraf.imarith(outmask, "*", med5, med5) if i > 1: delete(_output) delete(_output) iraf.imutil.imexpr(expr="(1.-b)*a+c", a=oldoutput, b=outmask, c=med5, output=_output, verbose='no') # cleanup and get ready for next iteration delete(oldoutput) iraf.imcopy(_output, oldoutput, verbose='no') if npix == 0: stop = 'yes' i = i + 1 if i > niter: stop = 'yes' # delete temp files delete(blk + "," + lapla + "," + deriv2 + "," + med5) delete(med3 + "," + med7 + "," + noise + "," + sigmap) delete(firstsel + "," + starreject + "," + gfirstsel) delete(finalsel + "," + inputmask) if skyval > 0: iraf.imarith(_output, "-", skyval, _output) delete('_kernel' + "," + '_gkernel') delete(oldoutput)
def lacos_im(_input, _output='clean.fits', outmask='mask.fits', gain=1.3, readn=9, xorder=9, yorder=9, sigclip=4.5, sigfrac=0.5, objlim=1, skyval=0, niter=2, verbose=True, interactive=False): # print "LOGX:: Entering `lacos_im` method/function in %(__file__)s" % # globals() import ntt from ntt.util import delete import sys import re import os import string from pyraf import iraf import numpy as np iraf.convolve.bilinear = 'no' iraf.convolve.radsym = 'no' # make temporary files oldoutput, galaxy, skymod, med5 = 'oldoutput.fits', 'galaxy.fits', 'skymod.fits', 'med5.fits' blk, lapla, med3, med7, sub5, sigima, finalsel = 'blk.fits', 'lapla.fits', 'med3.fits', 'med7.fits', 'sub5.fits', 'sigima.fits', 'finalsel.fits' deriv2, noise, sigmap, firstsel, starreject = 'deriv2.fits', 'noise.fits', 'sigmap.fits', 'firstsel.fits', 'starreject.fits' inputmask, gfirstsel = 'inputmask.fits', 'gfirstsel.fits' f = open('_kernel', 'w') f.write('0 -1 0;\n-1 4 -1;\n0 -1 0') f.close() # create growth kernel f = open('_gkernel', 'w') f.write('1 1 1;\n1 1 1;\n1 1 1') f.close() gkernel = np.array([[1.0, 1.0, 1.0], [1.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) # initialize loop usegain = gain i = 1 stop = 'no' previous = 0 if not _output: _output = _input arrayinput, headerinput = ntt.cosmics.fromfits(_input, verbose=False) ntt.cosmics.tofits(outmask, np.float32( arrayinput - arrayinput), headerinput, verbose=False) delete(oldoutput) if skyval > 0: arrayoldoutput = arrayinput + skyval else: arrayoldoutput = arrayinput ntt.cosmics.tofits(oldoutput, np.float32( arrayoldoutput), headerinput, verbose=False) # start iterations while stop == 'no': # take second-order derivative (Laplacian) of input image # kernel is convolved with subsampled image, in order to remove negative # pattern around high pixels delete(blk) delete(lapla) delete(deriv2) iraf.blkrep(oldoutput, blk, 2, 2) iraf.convolve(blk, lapla, '_kernel') iraf.imreplace(lapla, 0, upper=0, lower='INDEF', radius=0) iraf.blkavg(lapla, deriv2, 2, 2, option="average") delete(med5) # create model of background flux - 5x5 box should exclude all CRs iraf.median(oldoutput, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imreplace(med5, 0.0001, upper=0, lower='INDEF', radius=0) # create noise model delete(noise) iraf.imutil.imexpr(expr='sqrt(a*' + str(usegain) + '+' + str(readn) + '**2)/' + str(usegain), a=med5, output=noise, verbose='no') # divide Laplacian by noise model delete(sigmap) iraf.imarith(deriv2, "/", noise, sigmap) # Laplacian of blkreplicated image counts edges twice: iraf.imarith(sigmap, "/", 2., sigmap) # removal of large structure (bright, extended objects) delete(med5) iraf.median(sigmap, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') arraysigmap, headersigmap = ntt.cosmics.fromfits(sigmap, verbose=False) arraymed5, headermed5 = ntt.cosmics.fromfits(med5, verbose=False) arraysigmap = arraysigmap - arraymed5 iraf.imarith(sigmap, "-", med5, sigmap) # find all candidate cosmic rays # this selection includes sharp features such as stars and HII regions delete(firstsel) iraf.imcopy(sigmap, firstsel, verbose='no') iraf.imreplace(firstsel, 0, upper=sigclip, lower='INDEF', radius=0) iraf.imreplace(firstsel, 1, lower=0.1, upper='INDEF', radius=0) # arraygfirst=arraysigmap # arraygfirst = np.where(arraygfirst<sigclip,0,arraygfirst) # arraygfirst = np.where(arraygfirst>0.1,1,arraygfirst) # compare candidate CRs to median filtered image # this step rejects bright, compact sources from the initial CR list # subtract background and smooth component of objects delete(med3) delete(med7) iraf.median(oldoutput, med3, 3, 3, zlo='INDEF', zhi='INDEF', verbose='no') iraf.median(med3, med7, 7, 7, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imarith(med3, "-", med7, med3) iraf.imarith(med3, "/", noise, med3) iraf.imreplace(med3, 0.01, upper=0.01, lower='INDEF', radius=0) # compare CR flux to object flux delete(starreject) iraf.imutil.imexpr(expr="(a*b)/c", a=firstsel, b=sigmap, c=med3, output=starreject, verbose='no') # discard if CR flux <= objlim * object flux iraf.imreplace(starreject, 0, upper=objlim, lower='INDEF', radius=0) iraf.imreplace(starreject, 1, lower=0.5, upper='INDEF', radius=0) iraf.imarith(firstsel, "*", starreject, firstsel) # grow CRs by one pixel and check in original sigma map delete(gfirstsel) iraf.convolve(firstsel, gfirstsel, '_gkernel') iraf.imreplace(gfirstsel, 1, lower=0.5, upper='INDEF', radius=0) iraf.imarith(gfirstsel, "*", sigmap, gfirstsel) iraf.imreplace(gfirstsel, 0, upper=sigclip, lower='INDEF', radius=0) iraf.imreplace(gfirstsel, 1, lower=0.1, upper='INDEF', radius=0) # grow CRs by one pixel and lower detection limit sigcliplow = sigfrac * sigclip delete(finalsel) iraf.convolve(gfirstsel, finalsel, '_gkernel') iraf.imreplace(finalsel, 1, lower=0.5, upper='INDEF', radius=0) iraf.imarith(finalsel, "*", sigmap, finalsel) iraf.imreplace(finalsel, 0, upper=sigcliplow, lower='INDEF', radius=0) iraf.imreplace(finalsel, 1, lower=0.1, upper='INDEF', radius=0) # determine number of CRs found in this iteration delete(gfirstsel) iraf.imutil.imexpr(expr="(1-b)*a", a=outmask, b=finalsel, output=gfirstsel, verbose='no') npix = iraf.imstat(gfirstsel, fields="npix", lower=0.5, upper='INDEF', Stdout=1) # create cleaned output image; use 3x3 median with CRs excluded delete(med5) iraf.imarith(outmask, "+", finalsel, outmask) iraf.imreplace(outmask, 1, lower=1, upper='INDEF', radius=0) delete(inputmask) iraf.imutil.imexpr(expr="(1-10000*a)", a=outmask, output=inputmask, verbose='no') iraf.imarith(oldoutput, "*", inputmask, inputmask) delete(med5) iraf.median(inputmask, med5, 5, 5, zloreject=- 9999, zhi='INDEF', verbose='no') iraf.imarith(outmask, "*", med5, med5) if i > 1: delete(_output) delete(_output) iraf.imutil.imexpr(expr="(1.-b)*a+c", a=oldoutput, b=outmask, c=med5, output=_output, verbose='no') # cleanup and get ready for next iteration delete(oldoutput) iraf.imcopy(_output, oldoutput, verbose='no') if npix == 0: stop = 'yes' i = i + 1 if i > niter: stop = 'yes' # delete temp files delete(blk + "," + lapla + "," + deriv2 + "," + med5) delete(med3 + "," + med7 + "," + noise + "," + sigmap) delete(firstsel + "," + starreject + "," + gfirstsel) delete(finalsel + "," + inputmask) if skyval > 0: iraf.imarith(_output, "-", skyval, _output) delete('_kernel' + "," + '_gkernel') delete(oldoutput)
def lacos(_input0, output='clean.fits', outmask='mask.fits', gain=1.3, readn=9, xorder=9, yorder=9, sigclip=4.5, sigfrac=0.5, objlim=1, verbose=True, interactive=False): # print "LOGX:: Entering `lacos` method/function in %(__file__)s" % # globals() import ntt from ntt.util import delete import sys import re import os import string from pyraf import iraf import numpy as np oldoutput, galaxy, skymod, med5 = 'oldoutput.fits', 'galaxy.fits', 'skymod.fits', 'med5.fits' blk, lapla, med3, med7, sub5, sigima, finalsel = 'blk.fits', 'lapla.fits', 'med3.fits', 'med7.fits', 'sub5.fits', 'sigima.fits', 'finalsel.fits' deriv2, noise, sigmap, firstsel, starreject = 'deriv2.fits', 'noise.fits', 'sigmap.fits', 'firstsel.fits', 'starreject.fits' inputmask = 'inputmask.fits' # set some parameters in standard IRAF tasks iraf.convolve.bilinear = 'no' iraf.convolve.radsym = 'no' # create Laplacian kernel # laplkernel = np.array([[0.0, -1.0, 0.0], [-1.0, 4.0, -1.0], [0.0, -1.0, 0.0]]) f = open('_kernel', 'w') f.write('0 -1 0;\n-1 4 -1;\n0 -1 0') f.close() # create growth kernel f = open('_gkernel', 'w') f.write('1 1 1;\n1 1 1;\n1 1 1') f.close() gkernel = np.array([[1.0, 1.0, 1.0], [1.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) delete(galaxy) delete(skymod) delete(oldoutput) if not output: output = _input0 else: delete(output) iraf.imcopy(_input0, output, verbose='no') delete('_xxx.fits,_yyy.fits') iraf.imcopy(_input0 + '[350:550,*]', '_xxx.fits', verbose='no') _input = '_xxx.fits' arrayinput, headerinput = ntt.cosmics.fromfits(_input, verbose=False) ntt.cosmics.tofits(outmask, np.float32( arrayinput - arrayinput), headerinput, verbose=False) # subtract object spectra if desired iraf.fit1d(_input, galaxy, "fit", axis=2, order=9, func="leg", low=4., high=4., nav=1, inter='no', sample="*", niter=3, grow=0, cursor="") iraf.imarith(_input, "-", galaxy, oldoutput) # Subtract sky lines iraf.fit1d(oldoutput, skymod, "fit", axis=1, order=5, func="leg", low=4., high=4., inter='no', sample="*", nav=1, niter=3, grow=0, cursor="") iraf.imarith(oldoutput, "-", skymod, oldoutput) arrayoldoutput, headeroldoutput = ntt.cosmics.fromfits( oldoutput, verbose=False) # add object spectra to sky model iraf.imarith(skymod, "+", galaxy, skymod) delete(med5) # add median of residuals to sky model iraf.median(oldoutput, med5, 5, 5, zlor='INDEF', zhir='INDEF', verbose='no') # m5 = ndimage.filters.median_filter(_inputarray, size=5, mode='mirror') iraf.imarith(skymod, "+", med5, med5) # take second-order derivative (Laplacian) of input image # kernel is convolved with subsampled image, in order to remove negative # pattern around high pixels delete(blk) delete(lapla) iraf.blkrep(oldoutput, blk, 2, 2) iraf.convolve(blk, lapla, '_kernel') iraf.imreplace(lapla, 0, upper=0, lower='INDEF') delete(deriv2) delete(noise) iraf.blkavg(lapla, deriv2, 2, 2, option="average") # create noise model iraf.imutil.imexpr(expr='sqrt(a*' + str(gain) + '+' + str(readn) + '**2)/' + str(gain), a=med5, output=noise, verbose='no') iraf.imreplace(med5, 0.00001, upper=0, lower='INDEF') # divide Laplacian by noise model delete(sigmap) iraf.imutil.imexpr(expr='(a/b)/2', a=deriv2, b=noise, output=sigmap, verbose='no') # removal of large structure (bright, extended objects) delete(med5) iraf.median(sigmap, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imarith(sigmap, "-", med5, sigmap) # find all candidate cosmic rays # this selection includes sharp features such as stars and HII regions arraysigmap, headersigmap = ntt.cosmics.fromfits(sigmap, verbose=False) arrayf = np.where(arraysigmap < sigclip, 0, arraysigmap) arrayf = np.where(arrayf > 0.1, 1, arrayf) ntt.cosmics.tofits(firstsel, np.float32( arrayf), headersigmap, verbose=False) # compare candidate CRs to median filtered image # this step rejects bright, compact sources from the initial CR list # subtract background and smooth component of objects delete(med3) iraf.median(oldoutput, med3, 3, 3, zlo='INDEF', zhi='INDEF', verbose='no') delete(med7) delete('_' + med3) iraf.median(med3, med7, 7, 7, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imutil.imexpr(expr='(a-b)/c', a=med3, b=med7, c=noise, output='_' + med3, verbose='no') iraf.imreplace('_' + med3, 0.01, upper=0.01, lower='INDEF') # compare CR flux to object flux delete(starreject) iraf.imutil.imexpr(expr='a+b+c', a=firstsel, b=sigmap, c="_" + med3, output=starreject, verbose='no') # discard if CR flux <= objlim * object flux iraf.imreplace(starreject, 0, upper=objlim, lower='INDEF') iraf.imreplace(starreject, 1, lower=0.5, upper='INDEF') iraf.imarith(firstsel, "*", starreject, firstsel) # grow CRs by one pixel and check in original sigma map arrayfirst, headerfirst = ntt.cosmics.fromfits(firstsel, verbose=False) arraygfirst = ntt.cosmics.my_convolve_with_FFT2(arrayfirst, gkernel) arraygfirst = np.where(arraygfirst > 0.5, 1, arraygfirst) arraygfirst = arraygfirst * arraysigmap arraygfirst = np.where(arraygfirst < sigclip, 0, arraygfirst) arraygfirst = np.where(arraygfirst > 0.1, 1, arraygfirst) # grow CRs by one pixel and lower detection limit sigcliplow = sigfrac * sigclip # Finding neighbouring pixels affected by cosmic rays arrayfinal = ntt.cosmics.my_convolve_with_FFT2(arraygfirst, gkernel) arrayfinal = np.where(arrayfinal > 0.5, 1, arrayfinal) arrayfinal = arrayfinal * arraysigmap arrayfinal = np.where(arrayfinal < sigcliplow, 0, arrayfinal) arrayfinal = np.where(arrayfinal > 0.1, 1, arrayfinal) # determine number of CRs found in this iteration arraygfirst = (1 - (arrayfinal - arrayfinal)) * arrayfinal npix = [str(int(np.size(np.where(arraygfirst > 0.5)) / 2.))] # create cleaned output image; use 3x3 median with CRs excluded arrayoutmask = np.where(arrayfinal > 1, 1, arrayfinal) ntt.cosmics.tofits(outmask, np.float32( arrayoutmask), headerfirst, verbose=False) delete(inputmask) arrayinputmask = (1 - (10000 * arrayoutmask)) * arrayoldoutput ntt.cosmics.tofits(inputmask, np.float32( arrayinputmask), headerfirst, verbose=False) delete(med5) iraf.median(inputmask, med5, 5, 5, zloreject=- 9999, zhi='INDEF', verbose='no') iraf.imarith(outmask, "*", med5, med5) delete('_yyy.fits') iraf.imutil.imexpr(expr='(1-a)*b+c', a=outmask, b=oldoutput, c=med5, output='_yyy.fits', verbose='no') # add sky and object spectra back in iraf.imarith('_yyy.fits', "+", skymod, '_yyy.fits') # cleanup and get ready for next iteration if npix == 0: stop = yes # delete temp files iraf.imcopy('_yyy.fits', output + '[350:550,*]', verbose='no') delete(blk + "," + lapla + "," + deriv2 + "," + med5) delete(med3 + "," + med7 + "," + noise + "," + sigmap) delete(firstsel + "," + starreject) delete(finalsel + "," + inputmask) delete(oldoutput + "," + skymod + "," + galaxy) delete("_" + med3 + ",_" + sigmap) delete('_kernel' + "," + '_gkernel') delete(outmask) delete('_xxx.fits,_yyy.fits')
def lacos_spec(_input, output='clean.fits', outmask='mask.fits', gain=1.3, readn=9,\ xorder=9, yorder=3, sigclip=4.5, sigfrac=0.5, objlim=1, niter=4, instrument='kastr', verbose=True, interactive=False): # print "LOGX:: Entering `lacos` method/function in %(__file__)s" % # globals() import lickshane import sys import re import os import string from pyraf import iraf import numpy as np oldoutput, galaxy, skymod, med5 = 'oldoutput.fits', 'galaxy.fits', 'skymod.fits', 'med5.fits' blk, lapla, med3, med7, sub5, sigima, finalsel = 'blk.fits', 'lapla.fits', 'med3.fits', 'med7.fits', 'sub5.fits', 'sigima.fits', 'finalsel.fits' deriv2, noise, sigmap, firstsel, starreject = 'deriv2.fits', 'noise.fits', 'sigmap.fits', 'firstsel.fits', 'starreject.fits' inputmask = 'inputmask.fits' # set some parameters in standard IRAF tasks iraf.convolve.bilinear = 'no' iraf.convolve.radsym = 'no' # create Laplacian kernel # laplkernel = np.array([[0.0, -1.0, 0.0], [-1.0, 4.0, -1.0], [0.0, -1.0, 0.0]]) f = open('_kernel', 'w') f.write('0 -1 0;\n-1 4 -1;\n0 -1 0') f.close() # create growth kernel f = open('_gkernel', 'w') f.write('1 1 1;\n1 1 1;\n1 1 1') f.close() gkernel = np.array([[1.0, 1.0, 1.0], [1.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) lickshane.util.delete(galaxy) lickshane.util.delete(skymod) lickshane.util.delete(oldoutput) if not output: output = _input else: os.system('cp ' + _input + ' ' + output) os.system('cp ' + _input + ' ' + oldoutput) arrayinput, headerinput = lickshane.cosmics.fromfits(oldoutput, verbose=False) lickshane.cosmics.tofits(outmask, np.float32(arrayinput - arrayinput), headerinput, verbose=False) if instrument in ['kastr']: axis1 = 1 axis2 = 2 elif instrument in ['kastb']: axis1 = 2 axis2 = 1 # subtract object spectra if desired if xorder > 0: iraf.fit1d(oldoutput, galaxy, "fit", axis=axis1, order=xorder, func="leg", low=4., high=4., nav=1, inter='no', sample="*", niter=3, grow=0, cursor="") iraf.imarith(oldoutput, "-", galaxy, oldoutput) else: lickshane.cosmics.tofits(galaxy, np.float32(arrayinput - arrayinput), headerinput, verbose=False) # Subtract sky lines if yorder > 0: iraf.fit1d(oldoutput, skymod, "fit", axis=axis2, order=yorder, func="leg", low=4., high=4., inter='no', sample="*", nav=1, niter=3, grow=0, cursor="") iraf.imarith(oldoutput, "-", skymod, oldoutput) else: lickshane.cosmics.tofits(skymod, np.float32(arrayinput - arrayinput), headerinput, verbose=False) arrayoldoutput, headeroldoutput = lickshane.cosmics.fromfits(oldoutput, verbose=False) # add object spectra to sky model iraf.imarith(skymod, "+", galaxy, skymod) ########### ## start iteration ########### ii = 0 while ii < niter: print ii # add median of residuals to sky model lickshane.util.delete(med5) iraf.median(oldoutput, med5, 5, 5, zlor='INDEF', zhir='INDEF', verbose='no') # m5 = ndimage.filters.median_filter(_inputarray, size=5, mode='mirror') iraf.imarith(skymod, "+", med5, med5) # take second-order derivative (Laplacian) of input image # kernel is convolved with subsampled image, in order to remove negative # pattern around high pixels lickshane.util.delete(blk) lickshane.util.delete(lapla) lickshane.util.delete(deriv2) lickshane.util.delete(noise) lickshane.util.delete(sigmap) iraf.blkrep(oldoutput, blk, 2, 2) iraf.convolve(blk, lapla, '_kernel') iraf.imreplace(lapla, 0, upper=0, lower='INDEF') iraf.blkavg(lapla, deriv2, 2, 2, option="average") # create noise model iraf.imutil.imexpr(expr='sqrt(a*' + str(gain) + '+' + str(readn) + '**2)/' + str(gain), a=med5, output=noise, verbose='no') iraf.imreplace(med5, 0.00001, upper=0, lower='INDEF') # divide Laplacian by noise model iraf.imutil.imexpr(expr='(a/b)/2', a=deriv2, b=noise, output=sigmap, verbose='no') # removal of large structure (bright, extended objects) lickshane.util.delete(med5) iraf.median(sigmap, med5, 5, 5, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imarith(sigmap, "-", med5, sigmap) # find all candidate cosmic rays # this selection includes sharp features such as stars and HII regions arraysigmap, headersigmap = lickshane.cosmics.fromfits(sigmap, verbose=False) arrayf = np.where(arraysigmap < sigclip, 0, arraysigmap) arrayf = np.where(arrayf > 0.1, 1, arrayf) lickshane.cosmics.tofits(firstsel, np.float32(arrayf), headersigmap, verbose=False) # compare candidate CRs to median filtered image # this step rejects bright, compact sources from the initial CR list # subtract background and smooth component of objects lickshane.util.delete(med3) iraf.median(oldoutput, med3, 3, 3, zlo='INDEF', zhi='INDEF', verbose='no') lickshane.util.delete(med7) lickshane.util.delete('_' + med3) iraf.median(med3, med7, 7, 7, zlo='INDEF', zhi='INDEF', verbose='no') iraf.imutil.imexpr(expr='(a-b)/c', a=med3, b=med7, c=noise, output='_' + med3, verbose='no') iraf.imreplace('_' + med3, 0.01, upper=0.01, lower='INDEF') # compare CR flux to object flux lickshane.util.delete(starreject) iraf.imutil.imexpr(expr='(a*b)/c', a=firstsel, b=sigmap, c="_" + med3, output=starreject, verbose='no') # ###### ##### ###### ##### ###### FOUND A BUG ? # iraf.imutil.imexpr(expr='a+b+c', a=firstsel, b=sigmap, # c="_" + med3, output=starreject, verbose='no') # discard if CR flux <= objlim * object flux iraf.imreplace(starreject, 0, upper=objlim, lower='INDEF') iraf.imreplace(starreject, 1, lower=0.5, upper='INDEF') iraf.imarith(firstsel, "*", starreject, firstsel) # grow CRs by one pixel and check in original sigma map arrayfirst, headerfirst = lickshane.cosmics.fromfits(firstsel, verbose=False) arraygfirst = lickshane.cosmics.my_convolve_with_FFT2( arrayfirst, gkernel) arraygfirst = np.where(arraygfirst > 0.5, 1, arraygfirst) arraygfirst = arraygfirst * arraysigmap arraygfirst = np.where(arraygfirst < sigclip, 0, arraygfirst) arraygfirst = np.where(arraygfirst > 0.1, 1, arraygfirst) # grow CRs by one pixel and lower detection limit sigcliplow = sigfrac * sigclip # Finding neighbouring pixels affected by cosmic rays arrayfinal = lickshane.cosmics.my_convolve_with_FFT2( arraygfirst, gkernel) arrayfinal = np.where(arrayfinal > 0.5, 1, arrayfinal) arrayfinal = arrayfinal * arraysigmap arrayfinal = np.where(arrayfinal < sigcliplow, 0, arrayfinal) arrayfinal = np.where(arrayfinal > 0.1, 1, arrayfinal) # determine number of CRs found in this iteration arraygfirst = (1 - (arrayfinal - arrayfinal)) * arrayfinal npix = [str(int(np.size(np.where(arraygfirst > 0.5)) / 2.))] # create cleaned output image; use 3x3 median with CRs excluded arrayoutmask = np.where(arrayfinal > 1, 1, arrayfinal) lickshane.cosmics.tofits(outmask, np.float32(arrayoutmask), headerfirst, verbose=False) lickshane.util.delete(inputmask) arrayinputmask = (1 - (10000 * arrayoutmask)) * arrayoldoutput lickshane.cosmics.tofits(inputmask, np.float32(arrayinputmask), headerfirst, verbose=False) lickshane.util.delete(med5) iraf.median(inputmask, med5, 5, 5, zloreject=-9999, zhi='INDEF', verbose='no') iraf.imarith(outmask, "*", med5, med5) lickshane.util.delete(output) iraf.imutil.imexpr(expr='(1-a)*b+c', a=outmask, b=oldoutput, c=med5, output=output, verbose='no') lickshane.util.delete(oldoutput) os.system('cp ' + output + ' ' + oldoutput) # add sky and object spectra back in iraf.imarith(output, "+", skymod, output) # cleanup and get ready for next iteration ii = ii + 1 if npix == 0: ii = niter # delete temp files lickshane.util.delete(blk + "," + lapla + "," + deriv2 + "," + med5) lickshane.util.delete(med3 + "," + med7 + "," + noise + "," + sigmap) lickshane.util.delete(firstsel + "," + starreject) lickshane.util.delete(finalsel + "," + inputmask) lickshane.util.delete(oldoutput + "," + skymod + "," + galaxy) lickshane.util.delete("_" + med3 + ",_" + sigmap) lickshane.util.delete('_kernel' + "," + '_gkernel') lickshane.util.delete(outmask)