def spatial_smoothing(imgseq, macro_pixel_dim):
images_reduced = measure.block_reduce(imgseq.as_array(),
block_size=(1, macro_pixel_dim, macro_pixel_dim),
func=np.nanmean,
cval=np.nan) #np.nanmedian(imgseq.as_array()))
dim_t, dim_x, dim_y = images_reduced.shape
imgseq_reduced = neo.ImageSequence(images_reduced,
units=imgseq.units,
spatial_scale=imgseq.spatial_scale * macro_pixel_dim,
macro_pixel_dim=macro_pixel_dim,
sampling_rate=imgseq.sampling_rate,
file_origin=imgseq.file_origin,
t_start=imgseq.t_start)
if 'array_annotations' in imgseq.annotations:
del imgseq.annotations['array_annotations']
imgseq_reduced.annotations.update(imgseq.annotations)
imgseq_reduced.name = imgseq.name + " "
imgseq_reduced.annotations.update(macro_pixel_dim=macro_pixel_dim)
imgseq_reduced.description = imgseq.description + \
"spatially downsampled ({}).".format(os.path.basename(__file__))
return imgseq_reduced
def spatial_smoothing(images, macro_pixel_dim):
# Now we need to reduce the noise from the images by performing a spatial smoothing
images_reduced = measure.block_reduce(
images, (1, macro_pixel_dim, macro_pixel_dim),
np.nanmean,
cval=np.nanmedian(images))
dim_t, dim_x, dim_y = images_reduced.shape
new_annotations = images.annotations.copy()
new_annotations['array_annotations'] = {
'x_coords': [i % dim_x for i in range(dim_x * dim_y)],
'y_coords': [i // dim_y for i in range(dim_x * dim_y)]
}
imgseq_reduced = neo.ImageSequence(images_reduced,
units=images.units,
spatial_scale=images.spatial_scale *
macro_pixel_dim,
sampling_rate=images.sampling_rate,
file_origin=images.file_origin,
annotations=new_annotations)
imgseq_reduced.name = images.name + " "
imgseq_reduced.annotations.update(macro_pixel_dim=macro_pixel_dim)
imgseq_reduced.description = images.description + "spatially downsampled ({}).".format(
os.path.basename(__file__))
return imgseq_reduced
def AnalogSignal2ImageSequence(block):
for seg_count, segment in enumerate(block.segments):
block.segments[seg_count].imagesequences = []
for asig_count, asig in enumerate(segment.analogsignals):
asig_array = asig.as_array()
dim_t, dim_channels = asig_array.shape
if 'x_coords' not in asig.array_annotations\
or 'y_coords' not in asig.array_annotations:
print('AnalogSignal {} in Segment {} has no spatial Information '\
.format(asig_count, seg_count)\
+ ' as array_annotations "x_coords" "y_coords", skip.')
break
coords = np.array(
[(x, y) for x, y in zip(asig.array_annotations['x_coords'],
asig.array_annotations['y_coords'])],
dtype=float)
if len(coords) != dim_channels:
raise IndexError("Number of channels doesn't agree with "\
+ "number of coordinates!")
dim_x = np.max(asig.array_annotations['x_coords']) + 1
dim_y = np.max(asig.array_annotations['y_coords']) + 1
image_data = np.empty((dim_t, dim_x, dim_y), dtype=asig.dtype)
image_data[:] = np.nan
for channel in range(dim_channels):
x, y = coords[channel]
x, y = int(x), int(y)
image_data[:, x, y] = asig_array[:, channel]
spatial_scale = asig.annotations['spatial_scale']
# array_annotations = {}
# for k, v in asig.array_annotations.items():
# array_annotations[k] = v.reshape((dim_x, dim_y))
imgseq = neo.ImageSequence(
image_data=image_data,
units=asig.units,
dtype=asig.dtype,
# t_start=asig.t_start, # NotImplementedError
sampling_rate=asig.sampling_rate,
name=asig.name,
description=asig.description,
file_origin=asig.file_origin,
# array_annotations=array_annotations,
**asig.annotations)
block.segments[seg_count].imagesequences.append(imgseq)
return block
def AnalogSignal2ImageSequence(block):
# ToDo: map 1D array annotations to 2D and update
for seg_count, segment in enumerate(block.segments):
for asig in segment.analogsignals:
asig_array = asig.as_array()
dim_t, dim_channels = asig_array.shape
# coords = asig.channel_index.coordinates
# temporary replacement
coords = np.array(
[(x, y) for x, y in zip(asig.array_annotations['x_coords'],
asig.array_annotations['y_coords'])],
dtype=float)
#
# spatial_scale = asig.annotations['spatial_scale']
# int_coords = np.round(np.array(coords)/spatial_scale).astype(int)
# print(int_coords)
if len(coords) != dim_channels:
raise IndexError("Number of channels doesn't agree with "\
+ "number of coordinates!")
dim_x, dim_y = determine_dims(coords)
image_data = np.empty((dim_t, dim_x, dim_y))
image_data[:] = np.nan
for channel in range(dim_channels):
x, y = coords[channel]
x, y = int(x), int(y)
image_data[:, x, y] = asig_array[:, channel]
# spatial_scale = determine_spatial_scale(coords)*coords.units
spatial_scale = asig.annotations['spatial_scale']
array_annotations = {}
for k, v in asig.array_annotations.items():
array_annotations[k] = v.reshape((dim_x, dim_y))
imgseq = neo.ImageSequence(
image_data=image_data,
units=asig.units,
sampling_rate=asig.sampling_rate,
name=asig.name,
description=asig.description,
file_origin=asig.file_origin,
# array_annotations=array_annotations,
**asig.annotations)
block.segments[seg_count].imagesequences.append(imgseq)
return block
def spatial_smoothing(images, macro_pixel_dim):
# Now we need to reduce the noise from the images by performing a spatial smoothing
#images_reduced = measure.block_reduce(images, (1, macro_pixel_dim, macro_pixel_dim), np.nanmean, cval = np.nanmedian(images))
images_reduced = scipy.signal.decimate(images,
macro_pixel_dim,
n=2,
ftype='fir',
axis=1,
zero_phase=True)
images_reduced = scipy.signal.decimate(images_reduced,
macro_pixel_dim,
n=2,
ftype='fir',
axis=2,
zero_phase=True)
dim_t, dim_x, dim_y = images_reduced.shape
imgseq_reduced = neo.ImageSequence(
images_reduced,
#units=images.units,
#spatial_scale=images.spatial_scale * macro_pixel_dim,
#sampling_rate=images.sampling_rate,
#file_origin=images.file_origin)#,
#**imgseq.annotations)
units=imgseq.units,
spatial_scale=imgseq.spatial_scale * macro_pixel_dim,
sampling_rate=imgseq.sampling_rate,
file_origin=imgseq.file_origin,
t_start=imgseq.t_start)
if 'array_annotations' in imgseq.annotations:
del imgseq.annotations['array_annotations']
imgseq_reduced.annotations.update(imgseq.annotations)
imgseq_reduced.name = imgseq.name + " "
imgseq_reduced.annotations.update(macro_pixel_dim=macro_pixel_dim)
imgseq_reduced.description = imgseq.description + \
"spatially downsampled ({}).".format(os.path.basename(__file__))
return imgseq_reduced
# re-converting into analogsignal
signal = activity.reshape((dim_t, dim_x * dim_y))
asig = block.segments[0].analogsignals[0].duplicate_with_new_data(signal)
asig.array_annotate(**block.segments[0].analogsignals[0].array_annotations)
asig.name += ""
asig.description += "Deconvoluted activity using the given {} kernel"\
.format(args.kernel)
block.segments[0].analogsignals[0] = asig
"""
# New method, creating a new block
# create an ImageSequence with the deconvoluted matrix
imgseq_deconv = neo.ImageSequence(
activities,
units=block.segments[0].analogsignals[0].units,
sampling_rate=block.segments[0].analogsignals[0].sampling_rate,
spatial_scale=block.segments[0].imagesequences[0].spatial_scale,
name=block.segments[0].analogsignals[0].name,
description=block.segments[0].analogsignals[0].description)
# create a new Block & Segment and append the ImageSequence
segm_deconv = neo.Segment()
segm_deconv.annotations = block.segments[0].annotations
segm_deconv.annotate(kernel=args.kernel) #ToDo: parameters annotations
segm_deconv.imagesequences.append(imgseq_deconv)
block_deconv = neo.Block()
block_deconv.segments.append(segm_deconv)
block_deconv.name = block.name
block_deconv.description = block.description
block_deconv.annotations = block.annotations
max_Niter=args.max_Niter,
convergence_limit=args.convergence_limit,
kernelX=kernelX,
kernelY=kernelY,
kernelT=kernelT,
kernelHS=kernelHS)
if np.sum(args.gaussian_sigma):
vector_frames = smooth_frames(vector_frames, sigma=args.gaussian_sigma)
vec_imgseq = neo.ImageSequence(
vector_frames,
units='dimensionless',
dtype=complex,
# t_start=imgseq.t_start,
spatial_scale=imgseq.spatial_scale,
sampling_rate=imgseq.sampling_rate,
name='Optical Flow',
description='Horn-Schunck estimation of optical flow',
file_origin=imgseq.file_origin,
**imgseq.annotations)
if args.output_img is not None:
ax = plot_opticalflow(frames[0], vector_frames[0], skip_step=3)
ax.set_ylabel(f'pixel size: {imgseq.spatial_scale} '\
+ imgseq.spatial_scale.units.dimensionality.string)
# ax.set_xlabel('{:.3f} s'.format(imgseq.times[0].rescale('s')))
save_plot(args.output_img)
block.segments[0].imagesequences = [vec_imgseq]
block = ImageSequence2AnalogSignal(block)
['y_pos_sel']).T #array of 1000x 2500 (50x50) images
x_pos_sel = np.array(sio.loadmat(mat_fname)
['x_pos_sel']).T #array of 1000x 2500 (50x50) images
Piexel = x_pos_sel + y_pos_sel * 50
image_seq = np.empty([len(Signal[0]), 50, 50])
for t in range(len(Signal[0])):
# per ogni tempo
for px in range(len(y_pos_sel)):
# per ogni pixel
image_seq[t][x_pos_sel[px], y_pos_sel[px]] = Signal[px][t]
imageSequences = neo.ImageSequence(
image_seq,
sampling_rate=args.sampling_rate * pq.Hz,
spatial_scale=args.spatial_scale * pq.mm,
units='dimensionless')
# loading the data flips the images vertically!
#block = io.read_block()
block = neo.Block()
seg = neo.Segment(name='segment 0', index=0)
block.segments.append(seg)
print('vlock', block)
print('seg', block.segments[0])
block.segments[0].imagesequences.append(imageSequences)
# change data orientation to be top=ventral, right=lateral
def analogsignals_to_imagesequences(block):
# ToDo: map 1D array annotations to 2D and update
for seg_count, segment in enumerate(block.segments):
for asig_count, asig in enumerate(segment.analogsignals):
asig_array = asig.as_array()
dim_t, dim_channels = asig_array.shape
# coords = asig.channel_index.coordinates
# temporary replacement
if 'x_coords' not in asig.array_annotations\
or 'y_coords' not in asig.array_annotations:
print('AnalogSignal {} in Segment {} has no spatial Information '\
.format(asig_count, seg_count)\
+ ' as array_annotations "x_coords" "y_coords", skip.')
continue
coords = np.array(
[(x, y) for x, y in zip(asig.array_annotations['x_coords'],
asig.array_annotations['y_coords'])],
dtype=float)
#
# spatial_scale = asig.annotations['spatial_scale']
# int_coords = np.round(np.array(coords)/spatial_scale).astype(int)
# print(int_coords)
if len(coords) != dim_channels:
raise IndexError("Number of channels doesn't agree with "\
+ "number of coordinates!")
dim_x, dim_y = determine_dims(coords)
image_data = np.empty((dim_t, dim_x, dim_y), dtype=asig.dtype)
image_data[:] = np.nan
for channel in range(dim_channels):
x, y = coords[channel]
x, y = int(x), int(y)
image_data[:, x, y] = asig_array[:, channel]
# spatial_scale = determine_spatial_scale(coords)*coords.units
spatial_scale = asig.annotations['spatial_scale']
# array_annotations = {}
# for k, v in asig.array_annotations.items():
# array_annotations[k] = v.reshape((dim_x, dim_y))
imgseq = neo.ImageSequence(
image_data=image_data,
units=asig.units,
dtype=asig.dtype,
t_start=asig.t_start,
sampling_rate=asig.sampling_rate,
name=asig.name,
description=asig.description,
file_origin=asig.file_origin,
# array_annotations=array_annotations,
**asig.annotations)
imgseq.annotate(array_annotations=asig.array_annotations)
remove_annotations(imgseq, del_keys=['nix_name', 'neo_name'])
block.segments[seg_count].imagesequences.append(imgseq)
return block
