def run_fft_filter( files, bnd_file, modes, suffix= 'fft_frames/normed_frames/',
make_cubs=True, run_chomp=True ):
"""
For each frame in fdir, compute fft( frame ), then take the top
<modes> percent of the frequencies. Threshold these filtered
images at the mean (of pixels inside the cell boundary). Save the
thresholded image to fft_frames/ directory.
fdir : directory to frames
modes : percentage of modes (low->high) to use for low-pass filter
"""
# grab all frames, skip the directories
print "files", files
if os.path.isdir( files ):
fdir = files + '/'
dlist = os.listdir( fdir )
if os.uname()[0] == 'Linux':
frames = []
for f in dlist:
if f.endswith('npy') and not os.path.isdir( fdir+f ):
frames.append( f )
else:
frames = [ f for f in dlist if not os.path.isdir( fdir+f ) ]
else:
frames = [ files ]
fdir = files.rpartition( '/' )[0].rpartition( '/' )[0] +'/'
if not fdir.endswith( '/' ): fdir += '/'
savedir = fdir + suffix
# create the directory if necessary
make_dir( savedir )
# compute fft and filtered ifft on (normed) images
for frame in frames:
if frame.endswith( 'npy' ):
savename = frame.rstrip( '.npy' )
elif frame.endswith( 'txt' ):
savename = frame.rstrip( '.txt' )
else:
savename = frame
fft_data = {}
# store the modes
fft_data[ 'modes' ] = modes
try:
image = numpy.loadtxt( fdir+frame )
except ValueError:
image = numpy.load( fdir+frame )
except IOError:
raise
bnd = numpy.loadtxt( bnd_file )
# normalize the images and shift so centered at zero
image = normalize_image( image )
# FFT
X = fft_image( image )
# Y = H * X
Y = ideal_low_pass( X, r=modes )
Yinv = ifft_image( Y )
Ypow = log1p( numpy.abs( Yinv )**2 )
# crop the log of the low pass to get rid of the noise outside
# of the cell (zero out external points using boundary array)
crop_zeros = bnd * Ypow
crop_zeros.resize( Ypow.shape )
# append the cell with zeros cropped [-4]
#Y.append( crop_zeros )
m = crop_zeros[numpy.where( crop_zeros!=0 )].mean()
# everything above the mean, [-3]
#Ymean = numpy.ma.masked_less_equal( crop_zeros, m )
# just the boolean mask, [-2]
#Ymask = numpy.ma.masked_less_equal( crop_zeros, m ).mask
# now populate fft_data with the essentials and save to disk
#fft_data['fft'] = X
#fft_data['lowpass'] = Y
#fft_data['ifft'] = Yinv
#fft_data['fft_mag'] = Ypow
fft_data['ifft_nonzero'] = crop_zeros
fft_data['mean'] = m
# masking is easy, so no need to save.
#fft_data['ifft_mask_mean'] = Ymean
#fft_data['ifft_mask'] = Ymask
# create filename with chomp-readable name (remove the 'dot'
# from mode)
mode_part = str( modes ).partition( '.' )
mode_str = mode_part[0] + mode_part[2]
# split files into subdirectories
datadir = savedir + 'r'+mode_str
save_suffix = savename+'_r'+mode_str
make_dir( datadir )
dataname = datadir + '/' + savename+'_r'+mode_str+'.pkl'
# save our thresholded data for later analysis
save_fft( fft_data, dataname )
# additional work with chomp
if make_cubs:
fft2cub( dataname, datadir )
if run_chomp:
if not make_cubs:
pass
else:
cub_name = datadir +'_cub/' + save_suffix + '.cub'
cbetti_prefix = datadir + '_cbetti/' + save_suffix
betti_prefix = datadir + '_betti/' + save_suffix
make_dir( datadir + '_cbetti/' )
make_dir( datadir + '_betti/' )
chomp_betti.run_chomp( cub_name, cbetti_prefix+'.cbetti' )
chomp_betti.extract_betti( cbetti_prefix, betti_prefix )
def run_chomp( cub_name ):
"""
"""
save_prefix = cub_name.rstrip( '.cub' )
chomp_betti.run_chomp( cub_name, save_prefix+'.cbetti' )
chomp_betti.extract_betti( save_prefix )
#savedir = '/home/kellys/Dropbox/Shared-MSU-Kel/Research/RBC/'
#files = '/home/kellys/RBC_Deriv_120125_r01/'
#files = '/home/kellys/RBC_Deriv_120125_r005/'
#grab files, skip directories
if os.path.isdir (files):
fdir = files + '/'
dlist = os.listdir(fdir)
frames = []
for f in dlist:
if f.endswith('cbetti') and not os.path.isdir(fdir+f):
frames.append(f)
else:
frames = [ files ]
fdir = files.rpartition('/')[0].rpartition('/')[0]+'/'
#savedir = '/home/kellys/RBC_Deriv_120125_r01/'
#savedir = '/home/kellys/RBC_Deriv_110125_r01/'
print 'extracting from chomp Bettie'
frames.sort(key=R.natural_key)
k = 0
for frame in frames:
k = k+1
name = frame.rstrip('.cbetti')
CB.extract_betti(fdir+name, fdir+name)
barr = CB.read_betti_dir(fdir)
CB.plot_betti(barr, 120125, savedir, 1, None, 5000)