def plot_equalize():
X = array(sum([range(110, 121)] * 5 + [range(130, 141)] * 7, []))
Xc = cumsum(histogram(X, bins=30)[0])
Y = equalize(X)
Yc = cumsum(histogram(Y, bins=30)[0])
fig, ax = subplots()
ax.plot(Xc, c='b')
ax.plot(Yc, c='r')
show()
def plot_equalize(): X = array(sum([range(110, 121)] * 5 + [range(130, 141)] * 7, [])) Xc = cumsum(histogram(X, bins = 30)[0]) Y = equalize(X) Yc = cumsum(histogram(Y, bins = 30)[0]) fig, ax = subplots() ax.plot(Xc, c = 'b') ax.plot(Yc, c = 'r') show()
def test_equalize():
"""
Besides basic properties, does a linearity test (that may not be mathematically sound)
"""
X = array(sum([range(10, 21)] * 15 + [range(130, 141)] * 4, []))
Y = equalize(X, 255)
assert Y.min() == 0
assert Y.max() == 255
C = cumsum(histogram(Y, bins = 30)[0])
L = linspace(0, X.shape[0], 30)
err = sqrt(sum((L - C)**2)) / X.shape[0]
assert err < 0.5, 'equalized cdf appears to be not be linear ({0:.2f} rmse from linearity)'.format(err)
def test_equalize():
"""
Besides basic properties, does a linearity test (that may not be mathematically sound)
"""
X = array(sum([range(10, 21)] * 15 + [range(130, 141)] * 4, []))
Y = equalize(X, 255)
assert Y.min() == 0
assert Y.max() == 255
C = cumsum(histogram(Y, bins=30)[0])
L = linspace(0, X.shape[0], 30)
err = sqrt(sum((L - C)**2)) / X.shape[0]
assert err < 0.5, 'equalized cdf appears to be not be linear ({0:.2f} rmse from linearity)'.format(
err)
if 'shift' in fmt:
print 'shifting by {0}'.format(fmt['shift'])
column += fmt['shift']
if 'cut_gt' in fmt and 'cut_to' in fmt:
print 'cutting > {0:f} to {1:f} for {2:d}'.format(
fmt['cut_gt'], fmt['cut_to'], colnr)
cut_cnt += (column > fmt['cut_gt']).sum()
column[column > fmt['cut_gt']] = fmt['cut_to']
if 'cut_lt' in fmt and 'cut_to' in fmt:
print 'cutting < {0:f} to {1:f} for {2:d}'.format(
fmt['cut_lt'], fmt['cut_to'], colnr)
cut_cnt += (column < fmt['cut_lt']).sum()
column[column < fmt['cut_lt']] = fmt['cut_to']
if 'equalize' in fmt and fmt['equalize']:
print 'equalizing histogram'
column = equalize(column)
ax_after.hist(column, facecolor='blue', bins=30)
ax_after_cdf.hist(column, cumulative=True, facecolor='green', bins=30)
ax_frac.pie([
column.shape[0] - columnnonnum.shape[0] - cut_cnt,
columnnonnum.shape[0], cut_cnt
],
labels=['normal', 'NaN', 'cut'])
print 'frac', column.shape[0], [
column.shape[0] - columnnonnum.shape[0] - cut_cnt,
columnnonnum.shape[0], cut_cnt
]
show(block=False)
while True:
if 'cut_gt' in fmt:
print 'column {0:d} already has a cutoff at {1:d}'.format(
else:
fmt = copy(format['default'])
if 'shift' in fmt:
print 'shifting by {0}'.format(fmt['shift'])
column += fmt['shift']
if 'cut_gt' in fmt and 'cut_to' in fmt:
print 'cutting > {0:f} to {1:f} for {2:d}'.format(fmt['cut_gt'], fmt['cut_to'], colnr)
cut_cnt += (column > fmt['cut_gt']).sum()
column[column > fmt['cut_gt']] = fmt['cut_to']
if 'cut_lt' in fmt and 'cut_to' in fmt:
print 'cutting < {0:f} to {1:f} for {2:d}'.format(fmt['cut_lt'], fmt['cut_to'], colnr)
cut_cnt += (column < fmt['cut_lt']).sum()
column[column < fmt['cut_lt']] = fmt['cut_to']
if 'equalize' in fmt and fmt['equalize']:
print 'equalizing histogram'
column = equalize(column)
ax_after.hist(column, facecolor = 'blue', bins = 30)
ax_after_cdf.hist(column, cumulative = True, facecolor = 'green', bins = 30)
ax_frac.pie([column.shape[0] - columnnonnum.shape[0] - cut_cnt, columnnonnum.shape[0], cut_cnt], labels = ['normal', 'NaN', 'cut'])
print 'frac', column.shape[0], [column.shape[0] - columnnonnum.shape[0] - cut_cnt, columnnonnum.shape[0], cut_cnt]
show(block = False)
while True:
if 'cut_gt' in fmt:
print 'column {0:d} already has a cutoff at {1:d}'.format(colnr, fmt['cut_gt'])
close()
break
cut = raw_input('column {0:d} cutoff point? '.format(colnr))
if cut.strip() == '':
print 'no cutoff'
break
try:
