import numpy as np
from libertem.udf import UDF
[docs]class LogsumUDF(UDF):
"""
Sum up logscaled frames
In comparison to log-scaling the sum, this highlights regions with slightly higher
intensity that appear in many frames in relation to very high intensity in a few frames.
Examples
--------
>>> udf = LogsumUDF()
>>> result = ctx.run_udf(dataset=dataset, udf=udf)
>>> np.array(result["logsum"]).shape
(32, 32)
"""
def __init__(self):
super().__init__()
[docs] def get_backends(self):
return [
b for b in self.BACKEND_ALL
if b not in (
self.BACKEND_SCIPY_COO, self.BACKEND_SCIPY_CSR, self.BACKEND_SCIPY_CSC,
self.BACKEND_CUPY_SCIPY_COO, self.BACKEND_CUPY_SCIPY_CSC,
self.BACKEND_CUPY_SCIPY_CSR,
)
]
def get_result_buffers(self):
""
return {
'logsum': self.buffer(
kind='sig', dtype='float32', where='device'
),
}
def merge(self, dest, src):
""
dest.logsum[:] += src.logsum[:]
def merge_all(self, ordered_results):
''
chunks = [b.logsum for b in ordered_results.values()]
logsum = np.stack(chunks, axis=0).sum(axis=0)
return {'logsum': logsum}
def process_frame(self, frame):
""
self.results.logsum[:] += self.forbuf(
np.log(frame - np.min(frame) + 1),
self.results.logsum
)
def run_logsum(ctx, dataset, roi=None):
'''
Sum up logscaled frames
In comparison to log-scaling the sum, this highlights regions with slightly higher
intensity that appear in many frames in relation to very high intensity in a few frames.
Example:
f1 = (11, 101)
f2 = (11, 1)
f2 = (11, 1)
...
f10 = (11, 1)
log10(sum(f1 ... f10)) == (2.04, 2.04)
sum(log10(f1) ... log10(f10)) == (10.4, 2.04)
'''
udf = LogsumUDF()
return ctx.run_udf(dataset=dataset, udf=udf, roi=roi)