def centerParticles(self, oldstack, centerstack, badstack):
maxshift = self.params['maxshift']
centerparts = []
badparts = []
keeplist = []
i = 0
while partnum < numparts:
### if need more particles
### read 4000 parts from oldstack
### write centerparts to centerstack
### write badparts to badstack
### set current image
oldpart = oldparts[i]
### mirror about x
xmirror = numpy.flipud(oldpart)
### cross-correlate
xcc = correlator.cross_correlate(oldpart, xmirror)
### find peak
peakdict = peakfinder.findSubpixelPeak(xcc)
xpeak = correlator.wrap_coord(peakdict['pixel peak'], xcc.shape)
### mirror about y
ymirror = numpy.fliplr(oldpart)
### cross-correlate
ycc = correlator.cross_correlate(oldpart, ymirror)
### find peak
peakdict = peakfinder.findSubpixelPeak(ycc)
ypeak = correlator.wrap_coord(peakdict['pixel peak'], ycc.shape)
### mirror about y then x
xymirror = numpy.flipud(ymirror)
### cross-correlate
xycc = correlator.cross_correlate(oldpart, xymirror)
### find peak
peakdict = peakfinder.findSubpixelPeak(xycc)
xypeak = correlator.wrap_coord(peakdict['pixel peak'], xycc.shape)
### do some math to get shift
xshift = (ypeak[0] + xypeak[0]) / 4.0
yshift = (xpeak[0] + xypeak[0]) / 4.0
### shift particle, by integers only
if xshift < maxshift and yshift < maxshift:
xyshift = (xshift, yshift)
centerpart = ndimage.shift(oldpart,
shift=xyshift,
mode='wrap',
order=0)
centerparts.append(centerpart)
keeplist.append(partnum)
else:
badparts.append(oldpart)
return keeplist
def centerParticles(self, oldstack, centerstack, badstack): maxshift = self.params['maxshift'] centerparts = [] badparts = [] keeplist = [] i = 0 while partnum < numparts: ### if need more particles ### read 4000 parts from oldstack ### write centerparts to centerstack ### write badparts to badstack ### set current image oldpart = oldparts[i] ### mirror about x xmirror = numpy.flipud(oldpart) ### cross-correlate xcc = correlator.cross_correlate(oldpart, xmirror) ### find peak peakdict = peakfinder.findSubpixelPeak(xcc) xpeak = correlator.wrap_coord(peakdict['pixel peak'], xcc.shape) ### mirror about y ymirror = numpy.fliplr(oldpart) ### cross-correlate ycc = correlator.cross_correlate(oldpart, ymirror) ### find peak peakdict = peakfinder.findSubpixelPeak(ycc) ypeak = correlator.wrap_coord(peakdict['pixel peak'], ycc.shape) ### mirror about y then x xymirror = numpy.flipud(ymirror) ### cross-correlate xycc = correlator.cross_correlate(oldpart, xymirror) ### find peak peakdict = peakfinder.findSubpixelPeak(xycc) xypeak = correlator.wrap_coord(peakdict['pixel peak'], xycc.shape) ### do some math to get shift xshift = (ypeak[0] + xypeak[0])/4.0 yshift = (xpeak[0] + xypeak[0])/4.0 ### shift particle, by integers only if xshift < maxshift and yshift < maxshift: xyshift = (xshift, yshift) centerpart = ndimage.shift(oldpart, shift=xyshift, mode='wrap', order=0) centerparts.append(centerpart) keeplist.append(partnum) else: badparts.append(oldpart) return keeplist
def correlateTemplate(self):
## correlate
self.correlator.setImage(0, self.images['templateB'])
if self.settings['correlation type'] == 'phase':
cor = self.correlator.phaseCorrelate(zero=False)
else:
cor = self.correlator.crossCorrelate()
self.images['correlation'] = cor
# low pass filter
if self.settings['correlation lpf']:
cor = scipy.ndimage.gaussian_filter(
cor, self.settings['correlation lpf'])
# display correlation
self.setImage(cor, 'correlation')
# find peak
peakinfo = peakfinder.findSubpixelPeak(cor)
peakinfo['correlation'] = cor
peakinfo['templateB-unshifted'] = self.images['templateB-unshifted']
#self.printPeakInfo(peakinfo)
corpeak = peakinfo['subpixel peak']
ivtargets = [(corpeak[1], corpeak[0])]
self.setTargets(ivtargets, 'peak', block=True)
return peakinfo
def correlateTemplate(self):
## correlate
self.correlator.setImage(0, self.images["templateB"])
if self.settings["correlation type"] == "phase":
cor = self.correlator.phaseCorrelate(zero=False)
else:
cor = self.correlator.crossCorrelate()
self.images["correlation"] = cor
# low pass filter
if self.settings["correlation lpf"]:
cor = scipy.ndimage.gaussian_filter(cor, self.settings["correlation lpf"])
# display correlation
self.setImage(cor, "correlation")
# find peak
peakinfo = peakfinder.findSubpixelPeak(cor)
peakinfo["correlation"] = cor
peakinfo["templateB-unshifted"] = self.images["templateB-unshifted"]
# self.printPeakInfo(peakinfo)
corpeak = peakinfo["subpixel peak"]
ivtargets = [(corpeak[1], corpeak[0])]
self.setTargets(ivtargets, "peak", block=True)
return peakinfo
def correlateTemplate(self): ## correlate self.correlator.setImage(0, self.images['templateB']) if self.settings['correlation type'] == 'phase': cor = self.correlator.phaseCorrelate(zero=False) else: cor = self.correlator.crossCorrelate() self.images['correlation'] = cor # low pass filter if self.settings['correlation lpf']: cor = scipy.ndimage.gaussian_filter(cor, self.settings['correlation lpf']) # display correlation self.setImage(cor, 'correlation') # find peak peakinfo = peakfinder.findSubpixelPeak(cor) peakinfo['correlation'] = cor peakinfo['templateB-unshifted'] = self.images['templateB-unshifted'] #self.printPeakInfo(peakinfo) corpeak = peakinfo['subpixel peak'] ivtargets = [(corpeak[1], corpeak[0])] self.setTargets(ivtargets, 'peak', block=True) return peakinfo
def getShift(imgdata1, imgdata2):
# assumes images are square
print "Finding shift between", apDisplay.short(imgdata1["filename"]), "and", apDisplay.short(imgdata2["filename"])
dimension1 = imgdata1["camera"]["dimension"]["x"]
binning1 = imgdata1["camera"]["binning"]["x"]
dimension2 = imgdata2["camera"]["dimension"]["x"]
binning2 = imgdata2["camera"]["binning"]["x"]
finalsize = 512
# test to make sure images are at same mag
if imgdata1["scope"]["magnification"] != imgdata2["scope"]["magnification"]:
apDisplay.printWarning("Defocus pairs are at different magnifications, so shift can't be calculated.")
return None
# test to see if images capture the same area
if (dimension1 * binning1) != (dimension2 * binning2):
apDisplay.printWarning("Defocus pairs do not capture the same imaging area, so shift can't be calculated.")
return None
# images must not be less than finalsize (currently 512) pixels. This is arbitrary but for good reason
if dimension1 < finalsize or dimension2 < finalsize:
apDisplay.printWarning("Images must be greater than " + finalsize + " pixels to calculate shift.")
return None
# low pass filter 2 images to twice the final pixelsize BEFORE binning
shrinkfactor1 = dimension1 / finalsize
shrinkfactor2 = dimension2 / finalsize
binned1 = apImage.filterImg(imgdata1["image"], 1.0, shrinkfactor1 * 2)
binned2 = apImage.filterImg(imgdata2["image"], 1.0, shrinkfactor2 * 2)
# now bin 2 images
binned1 = apImage.binImg(binned1, shrinkfactor1)
binned2 = apImage.binImg(binned2, shrinkfactor2)
### fix for non-square images, correlation fails on non-square images
mindim = min(binned1.shape)
binned1 = binned1[:mindim, :mindim]
binned2 = binned2[:mindim, :mindim]
### use phase correlation, performs better than cross
pc = correlator.phase_correlate(binned1, binned2)
apImage.arrayToMrc(pc, "phaseCorrelate.mrc")
### find peak, filtering to 10.0 helps
peak = peakfinder.findSubpixelPeak(pc, lpf=10.0)
subpixpeak = peak["subpixel peak"]
shift = correlator.wrap_coord(subpixpeak, pc.shape)
peak["scalefactor"] = dimension2 / float(dimension1)
# print shift[0]*shrinkfactor1, shift[1]*shrinkfactor1
xshift = int(round(shift[0] * shrinkfactor1))
yshift = int(round(shift[1] * shrinkfactor1))
peak["shift"] = numpy.array((xshift, yshift))
apDisplay.printMsg("Determined shifts: %f %f" % (peak["shift"][0], peak["shift"][1]))
# print peak['shift']
# sys.exit(1)
return peak
def register2(image1, image2, angles):
im2filt = scipy.ndimage.spline_filter(image2)
peaks = []
for angle in angles:
image2 = scipy.ndimage.rotate(im2filt, angle, reshape=False)
#mrc.write(image2, 'rot.mrc')
pc = correlator.phase_correlate(image1, image2, zero=False)
mrc.write(pc, 'pc.mrc')
peak = peakfinder.findSubpixelPeak(pc)
result = (angle, peak['pixel peak value'], peak['subpixel peak value'], peak['snr'])
print result
peaks.append(result)
return peaks
def correlateOriginal(self, index, imagedata):
if index == 0:
self.originalimage = imagedata['image']
unbinned = {'row': 0.0, 'col': 0.0}
else:
## correlate
self.startTimer('scope change correlation')
try:
correlation_type = self.settings['correlation type']
except KeyError:
correlation_type = 'phase'
if correlation_type == 'cross':
cor = correlator.cross_correlate(self.originalimage,
imagedata['image'])
elif correlation_type == 'phase':
cor = correlator.phase_correlate(self.originalimage,
imagedata['image'],
False,
wiener=True)
else:
raise RuntimeError('invalid correlation type')
self.stopTimer('scope change correlation')
self.displayCorrelation(cor)
## find peak
self.startTimer('shift peak')
peak = peakfinder.findSubpixelPeak(cor)
self.stopTimer('shift peak')
self.logger.debug('Peak %s' % (peak, ))
pixelpeak = peak['subpixel peak']
self.startTimer('shift display')
self.displayPeak(pixelpeak)
self.stopTimer('shift display')
peakvalue = peak['subpixel peak value']
shift = correlator.wrap_coord(peak['subpixel peak'], cor.shape)
self.logger.debug('pixel shift (row,col): %s' % (shift, ))
## need unbinned result
binx = imagedata['camera']['binning']['x']
biny = imagedata['camera']['binning']['y']
unbinned = {'row': shift[0] * biny, 'col': shift[1] * binx}
shiftinfo = {
'previous': self.originalimage,
'next': imagedata,
'pixel shift': unbinned
}
return shiftinfo
def simpleCorrelation(array1,array2):
c = correlator.Correlator()
p = peakfinder.PeakFinder()
c.setImage(0,array1)
c.setImage(1,array2)
shape = array1.shape
corrimage = c.phaseCorrelate()
p.setImage(corrimage)
peak = peakfinder.findSubpixelPeak(corrimage,lpf=1.5)
shift = [0.0,0.0]
for i in (0,1):
if peak['subpixel peak'][i] > shape[i]/2:
shift[i] = peak['subpixel peak'][i] - shape[i]
else:
shift[i] = peak['subpixel peak'][i]
return shift
def simpleCorrelation(array1, array2):
c = correlator.Correlator()
p = peakfinder.PeakFinder()
c.setImage(0, array1)
c.setImage(1, array2)
shape = array1.shape
corrimage = c.phaseCorrelate()
p.setImage(corrimage)
peak = peakfinder.findSubpixelPeak(corrimage, lpf=1.5)
shift = [0.0, 0.0]
for i in (0, 1):
if peak['subpixel peak'][i] > shape[i] / 2:
shift[i] = peak['subpixel peak'][i] - shape[i]
else:
shift[i] = peak['subpixel peak'][i]
return shift
def correlateOriginal(self,index,imagedata):
if index == 0:
self.originalimage = imagedata['image']
unbinned = {'row':0.0, 'col': 0.0}
else:
## correlate
self.startTimer('scope change correlation')
try:
correlation_type = self.settings['correlation type']
except KeyError:
correlation_type = 'phase'
if correlation_type == 'cross':
cor = correlator.cross_correlate(self.originalimage,imagedata['image'])
elif correlation_type == 'phase':
cor = correlator.phase_correlate(self.originalimage,imagedata['image'],False,wiener=True)
else:
raise RuntimeError('invalid correlation type')
self.stopTimer('scope change correlation')
self.displayCorrelation(cor)
## find peak
self.startTimer('shift peak')
peak = peakfinder.findSubpixelPeak(cor)
self.stopTimer('shift peak')
self.logger.debug('Peak %s' % (peak,))
pixelpeak = peak['subpixel peak']
self.startTimer('shift display')
self.displayPeak(pixelpeak)
self.stopTimer('shift display')
peakvalue = peak['subpixel peak value']
shift = correlator.wrap_coord(peak['subpixel peak'], cor.shape)
self.logger.debug('pixel shift (row,col): %s' % (shift,))
## need unbinned result
binx = imagedata['camera']['binning']['x']
biny = imagedata['camera']['binning']['y']
unbinned = {'row':shift[0] * biny, 'col': shift[1] * binx}
shiftinfo = {'previous': self.originalimage, 'next': imagedata, 'pixel shift': unbinned}
return shiftinfo
def removeStageAlphaBacklash(self, tilts, preset_name, target, emtarget):
if len(tilts) < 2:
raise ValueError
## change to parent preset
try:
parentname = target['image']['preset']['name']
except:
adjust = False
else:
adjust = True
## acquire parent preset image, initial image
if adjust:
isoffset = self.getImageShiftOffset()
self.presetsclient.toScope(parentname)
self.setImageShiftOffset(isoffset)
imagedata0 = self.acquireCorrectedCameraImageData(0)
## tilt then return in slow increments
delta = math.radians(5.0)
n = 5
increment = delta / n
if tilts[1] - tilts[0] > 0:
sign = -1
else:
sign = 1
alpha = tilts[0] + sign * delta
self.instrument.tem.StagePosition = {'a': alpha}
time.sleep(1.0)
for i in range(n):
alpha -= sign * increment
self.instrument.tem.StagePosition = {'a': alpha}
time.sleep(1.0)
if adjust:
## acquire parent preset image, final image
imagedata1 = self.acquireCorrectedCameraImageData(1)
self.presetsclient.toScope(preset_name)
## return to tomography preset
if emtarget['movetype'] == 'image shift':
presetdata = self.presetsclient.getPresetFromDB(preset_name)
self.moveAndPreset(presetdata, emtarget)
else:
self.presetsclient.toScope(preset_name)
self.setImageShiftOffset(isoffset)
## find shift between image0, image1
pc = correlator.phase_correlate(imagedata0['image'],
imagedata1['image'], False)
peakinfo = peakfinder.findSubpixelPeak(pc, lpf=1.5)
subpixelpeak = peakinfo['subpixel peak']
shift = correlator.wrap_coord(subpixelpeak,
imagedata0['image'].shape)
shift = {'row': shift[0], 'col': shift[1]}
## transform pixel to image shift
oldscope = imagedata0['scope']
newscope = self.calclients['image shift'].transform(
shift, oldscope, imagedata0['camera'])
ishiftx = newscope['image shift']['x'] - oldscope['image shift'][
'x']
ishifty = newscope['image shift']['y'] - oldscope['image shift'][
'y']
oldishift = self.instrument.tem.ImageShift
newishift = {
'x': oldishift['x'] + ishiftx,
'y': oldishift['y'] + ishifty
}
self.logger.info(
'adjusting imageshift after backlash: dx,dy = %s,%s' %
(ishiftx, ishifty))
self.instrument.tem.ImageShift = newishift
def removeStageAlphaBacklash(self, tilts, preset_name, target, emtarget):
if len(tilts) < 2:
raise ValueError
## change to parent preset
try:
parentname = target['image']['preset']['name']
except:
adjust = False
else:
adjust = True
## acquire parent preset image, initial image
if adjust:
isoffset = self.getImageShiftOffset()
self.presetsclient.toScope(parentname)
self.setImageShiftOffset(isoffset)
imagedata0 = self.acquireCorrectedCameraImageData(0)
## tilt then return in slow increments
delta = math.radians(5.0)
n = 5
increment = delta/n
if tilts[1] - tilts[0] > 0:
sign = -1
else:
sign = 1
alpha = tilts[0] + sign*delta
self.instrument.tem.StagePosition = {'a': alpha}
time.sleep(1.0)
for i in range(n):
alpha -= sign*increment
self.instrument.tem.StagePosition = {'a': alpha}
time.sleep(1.0)
if adjust:
## acquire parent preset image, final image
imagedata1 = self.acquireCorrectedCameraImageData(1)
self.presetsclient.toScope(preset_name)
## return to tomography preset
if emtarget['movetype'] == 'image shift':
presetdata = self.presetsclient.getPresetFromDB(preset_name)
self.moveAndPreset(presetdata, emtarget)
else:
self.presetsclient.toScope(preset_name)
self.setImageShiftOffset(isoffset)
## find shift between image0, image1
pc = correlator.phase_correlate(imagedata0['image'], imagedata1['image'], False)
peakinfo = peakfinder.findSubpixelPeak(pc, lpf=1.5)
subpixelpeak = peakinfo['subpixel peak']
shift = correlator.wrap_coord(subpixelpeak, imagedata0['image'].shape)
shift = {'row': shift[0], 'col': shift[1]}
## transform pixel to image shift
oldscope = imagedata0['scope']
newscope = self.calclients['image shift'].transform(shift, oldscope, imagedata0['camera'])
ishiftx = newscope['image shift']['x'] - oldscope['image shift']['x']
ishifty = newscope['image shift']['y'] - oldscope['image shift']['y']
oldishift = self.instrument.tem.ImageShift
newishift = {'x': oldishift['x'] + ishiftx, 'y': oldishift['y'] + ishifty}
self.logger.info('adjusting imageshift after backlash: dx,dy = %s,%s' % (ishiftx,ishifty))
self.instrument.tem.ImageShift = newishift
def getShift(imgdata1, imgdata2):
#assumes images are square
print "Finding shift between", apDisplay.short(
imgdata1['filename']), "and", apDisplay.short(imgdata2['filename'])
dimension1 = imgdata1['camera']['dimension']['x']
binning1 = imgdata1['camera']['binning']['x']
dimension2 = imgdata2['camera']['dimension']['x']
binning2 = imgdata2['camera']['binning']['x']
finalsize = 512
#test to make sure images are at same mag
if imgdata1['scope']['magnification'] != imgdata2['scope']['magnification']:
apDisplay.printWarning(
"Defocus pairs are at different magnifications, so shift can't be calculated."
)
return None
#test to see if images capture the same area
if (dimension1 * binning1) != (dimension2 * binning2):
apDisplay.printWarning(
"Defocus pairs do not capture the same imaging area, so shift can't be calculated."
)
return None
#images must not be less than finalsize (currently 512) pixels. This is arbitrary but for good reason
if dimension1 < finalsize or dimension2 < finalsize:
apDisplay.printWarning("Images must be greater than " + finalsize +
" pixels to calculate shift.")
return None
#low pass filter 2 images to twice the final pixelsize BEFORE binning
shrinkfactor1 = dimension1 / finalsize
shrinkfactor2 = dimension2 / finalsize
binned1 = apImage.filterImg(imgdata1['image'], 1.0, shrinkfactor1 * 2)
binned2 = apImage.filterImg(imgdata2['image'], 1.0, shrinkfactor2 * 2)
#now bin 2 images
binned1 = apImage.binImg(binned1, shrinkfactor1)
binned2 = apImage.binImg(binned2, shrinkfactor2)
### fix for non-square images, correlation fails on non-square images
mindim = min(binned1.shape)
binned1 = binned1[:mindim, :mindim]
binned2 = binned2[:mindim, :mindim]
### use phase correlation, performs better than cross
pc = correlator.phase_correlate(binned1, binned2)
apImage.arrayToMrc(pc, "phaseCorrelate.mrc")
### find peak, filtering to 10.0 helps
peak = peakfinder.findSubpixelPeak(pc, lpf=10.0)
subpixpeak = peak['subpixel peak']
shift = correlator.wrap_coord(subpixpeak, pc.shape)
peak['scalefactor'] = dimension2 / float(dimension1)
#print shift[0]*shrinkfactor1, shift[1]*shrinkfactor1
xshift = int(round(shift[0] * shrinkfactor1))
yshift = int(round(shift[1] * shrinkfactor1))
peak['shift'] = numpy.array((xshift, yshift))
apDisplay.printMsg("Determined shifts: %f %f" %
(peak['shift'][0], peak['shift'][1]))
#print peak['shift']
#sys.exit(1)
return peak
def getTiltedRotateShift(img1, img2, tiltdiff, angle=0, bin=1, msg=True):
"""
takes two images tilted
with respect to one another
and tries to find overlap
img1 (as numpy array)
img2 (as numpy array)
tiltdiff (in degrees)
negative, img1 is more compressed (tilted)
positive, img2 is more compressed (tilted)
"""
### untilt images by stretching and compressing
# choose angle s/t compressFactor = 1/stretchFactor
# this only works if one image is untilted (RCT) of both images are opposite tilt (OTR)
#halftilt = abs(tiltdiff)/2.0
halftiltrad = math.acos(math.sqrt(math.cos(abs(tiltdiff)/180.0*math.pi)))
# go from zero tilt to half tilt
compressFactor = math.cos(halftiltrad)
# go from max tilt to half tilt
stretchFactor = math.cos(halftiltrad) / math.cos(abs(tiltdiff)/180.0*math.pi)
if tiltdiff > 0:
if msg is True:
apDisplay.printMsg("compress image 1")
untilt1 = transformImage(img1, compressFactor, angle)
untilt2 = transformImage(img2, stretchFactor, angle)
xfactor = compressFactor
else:
if msg is True:
apDisplay.printMsg("stretch image 1")
untilt1 = transformImage(img1, stretchFactor, angle)
untilt2 = transformImage(img2, compressFactor, angle)
xfactor = stretchFactor
### filtering was done earlier
filt1 = untilt1
filt2 = untilt2
if filt1.shape != filt2.shape:
newshape = ( max(filt1.shape[0],filt2.shape[0]), max(filt1.shape[1],filt2.shape[1]) )
apDisplay.printMsg("Resizing images to: "+str(newshape))
filt1 = apImage.frame_constant(filt1, newshape, filt1.mean())
filt2 = apImage.frame_constant(filt2, newshape, filt2.mean())
### cross-correlate
cc = correlator.cross_correlate(filt1, filt2, pad=True)
rad = min(cc.shape)/20.0
cc = apImage.highPassFilter(cc, radius=rad)
cc = apImage.normRange(cc)
cc = blackEdges(cc)
cc = apImage.normRange(cc)
cc = blackEdges(cc)
cc = apImage.normRange(cc)
cc = apImage.lowPassFilter(cc, radius=10.0)
#find peak
peakdict = peakfinder.findSubpixelPeak(cc, lpf=0)
#import pprint
#pprint.pprint(peak)
pixpeak = peakdict['subpixel peak']
if msg is True:
apDisplay.printMsg("Pixel peak: "+str(pixpeak))
apImage.arrayToJpegPlusPeak(cc, "guess-cross-ang"+str(abs(angle))+".jpg", pixpeak)
rawpeak = numpy.array([pixpeak[1], pixpeak[0]]) #swap coord
shift = numpy.asarray(correlator.wrap_coord(rawpeak, cc.shape))*bin
if msg is True:
apDisplay.printMsg("Found xy-shift btw two images"
+";\n\t SNR= "+str(round(peakdict['snr'],2))
+";\n\t halftilt= "+str(round(halftiltrad*180/math.pi, 3))
+";\n\t compressFactor= "+str(round(compressFactor, 3))
+";\n\t stretchFactor= "+str(round(stretchFactor, 3))
+";\n\t xFactor= "+str(round(xfactor, 3))
+";\n\t rawpeak= "+str(numpy.around(rawpeak*bin, 1))
+";\n\t shift= "+str(numpy.around(shift, 1))
)
return shift, xfactor, peakdict['snr']
def getTiltedRotateShift(img1, img2, tiltdiff, angle=0, bin=1, msg=True):
"""
takes two images tilted
with respect to one another
and tries to find overlap
img1 (as numpy array)
img2 (as numpy array)
tiltdiff (in degrees)
negative, img1 is more compressed (tilted)
positive, img2 is more compressed (tilted)
"""
### untilt images by stretching and compressing
# choose angle s/t compressFactor = 1/stretchFactor
# this only works if one image is untilted (RCT) of both images are opposite tilt (OTR)
#halftilt = abs(tiltdiff)/2.0
halftiltrad = math.acos(math.sqrt(math.cos(abs(tiltdiff)/180.0*math.pi)))
# go from zero tilt to half tilt
compressFactor = math.cos(halftiltrad)
# go from max tilt to half tilt
stretchFactor = math.cos(halftiltrad) / math.cos(abs(tiltdiff)/180.0*math.pi)
if tiltdiff > 0:
if msg is True:
apDisplay.printMsg("compress image 1")
untilt1 = transformImage(img1, compressFactor, angle)
untilt2 = transformImage(img2, stretchFactor, angle)
xfactor = compressFactor
else:
if msg is True:
apDisplay.printMsg("stretch image 1")
untilt1 = transformImage(img1, stretchFactor, angle)
untilt2 = transformImage(img2, compressFactor, angle)
xfactor = stretchFactor
### filtering was done earlier
filt1 = untilt1
filt2 = untilt2
if filt1.shape != filt2.shape:
newshape = ( max(filt1.shape[0],filt2.shape[0]), max(filt1.shape[1],filt2.shape[1]) )
apDisplay.printMsg("Resizing images to: "+str(newshape))
filt1 = apImage.frame_constant(filt1, newshape, filt1.mean())
filt2 = apImage.frame_constant(filt2, newshape, filt2.mean())
### cross-correlate
cc = correlator.cross_correlate(filt1, filt2, pad=True)
rad = min(cc.shape)/20.0
cc = apImage.highPassFilter(cc, radius=rad)
cc = apImage.normRange(cc)
cc = blackEdges(cc)
cc = apImage.normRange(cc)
cc = blackEdges(cc)
cc = apImage.normRange(cc)
cc = apImage.lowPassFilter(cc, radius=10.0)
#find peak
peakdict = peakfinder.findSubpixelPeak(cc, lpf=0)
#import pprint
#pprint.pprint(peak)
pixpeak = peakdict['subpixel peak']
if msg is True:
apDisplay.printMsg("Pixel peak: "+str(pixpeak))
apImage.arrayToJpegPlusPeak(cc, "guess-cross-ang"+str(abs(angle))+".jpg", pixpeak)
rawpeak = numpy.array([pixpeak[1], pixpeak[0]]) #swap coord
shift = numpy.asarray(correlator.wrap_coord(rawpeak, cc.shape))*bin
if msg is True:
apDisplay.printMsg("Found xy-shift btw two images"
+";\n\t SNR= "+str(round(peakdict['snr'],2))
+";\n\t halftilt= "+str(round(halftiltrad*180/math.pi, 3))
+";\n\t compressFactor= "+str(round(compressFactor, 3))
+";\n\t stretchFactor= "+str(round(stretchFactor, 3))
+";\n\t xFactor= "+str(round(xfactor, 3))
+";\n\t rawpeak= "+str(numpy.around(rawpeak*bin, 1))
+";\n\t shift= "+str(numpy.around(shift, 1))
)
return shift, xfactor, peakdict['snr']