def test_9x9_linear_reduce(self):
res = {
'weights': [0.2270270, 0.3162162, 0.0702702],
'offsets': [0, 1.3846153, 3.2307692]
}
this = kernel_binom_linear(kernel_binom(9, 2, 2))
self._compare_linear(res, this)
def test_9x9_reduce(self):
res = {
'weights': [0.2270270, 0.1945945, 0.1216216, 0.0540540, 0.0162162],
'offsets': [0, 1, 2, 3, 4]
}
this = kernel_binom(9, 2, 2)
self._compare(res, this)
def main():
parser = argparse.ArgumentParser(
description="Calculates gaussian kernel weights and offsets from a "
"binomial distribution and optionally adjust the weights "
"and offsets for a linearly-sampled gaussian blur shader.")
parser.add_argument("taps",
default=5,
type=int,
help="Specify the number of taps (kernel size)")
parser.add_argument(
"--expand",
default=0,
type=int,
help="How much to expand the tap count (expand outermost "
"coefficients)")
parser.add_argument(
"--reduce",
default=0,
type=int,
help="How many taps to discard at borders (eliminate outermost "
"coefficients)")
parser.add_argument(
"--linear",
default=False,
action='store_true',
help="Uses linear sampling to compute weights and offsets")
args = parser.parse_args()
taps = args.taps
exp = args.expand
red = args.reduce
linear = args.linear
print "Computing a", \
bold("{0}-tap".format(taps)), \
"filter kernel (+%(exp)s/-%(red)s) %(desc)s" % \
{
'taps': taps,
'exp': exp * 2,
'red': red * 2,
'desc': " (+linear reduction)" if linear else ""
}
print "Initial gaussian distribution: {0}".format(
bold(str(binom(taps - 1 + exp * 2))))
ntap = taps
res_discrete = kernel_binom(taps, exp, red)
res_linear = None
if linear:
res_linear = kernel_binom_linear(res_discrete)
ntap = (len(res_linear['weights']) * 2) - 1
if res_discrete is not None:
float_format = "{:.8f}"
print "Initial ", \
bold("{0}-tap".format(taps)), \
"filter kernel coefficients:".format(taps)
print "\tweights:", \
[float_format.format(x) for x in res_discrete["weights"]]
print "\toffsets:", \
[float_format.format(x) for x in res_discrete["offsets"]]
if linear and res_linear is not None:
print "\nOptimized", \
bold("{0}-tap".format(ntap)), \
"filter kernel coefficients:".format(ntap)
print "\tweights:", \
[float_format.format(x) for x in res_linear["weights"]]
print "\toffsets:", \
[float_format.format(x) for x in res_linear["offsets"]]
def test_9x9_linear_reduce(self):
res = {'weights': [0.2270270, 0.3162162, 0.0702702],
'offsets': [0, 1.3846153, 3.2307692]}
this = kernel_binom_linear(kernel_binom(9, 2, 2))
self._compare_linear(res, this)
def test_9x9_reduce(self):
res = {
'weights': [0.2270270, 0.1945945, 0.1216216, 0.0540540, 0.0162162],
'offsets': [0, 1, 2, 3, 4]}
this = kernel_binom(9, 2, 2)
self._compare(res, this)
def test_5x5_linear(self):
res = {'weights': [0.375, 0.3125],
'offsets': [0, 1.2]}
this = kernel_binom_linear(kernel_binom(5, 0, 0))
self._compare_linear(res, this)
def test_5x5(self):
res = {'weights': [0.375, 0.25, 0.0625],
'offsets': [0, 1, 2]}
this = kernel_binom(5, 0, 0)
self._compare(res, this)
def test_3x3(self):
res = {'weights': [0.5, 0.25],
'offsets': [0, 1]}
this = kernel_binom(3, 0, 0)
self._compare(res, this)
def test_5x5_linear(self):
res = {'weights': [0.375, 0.3125], 'offsets': [0, 1.2]}
this = kernel_binom_linear(kernel_binom(5, 0, 0))
self._compare_linear(res, this)
def test_5x5(self):
res = {'weights': [0.375, 0.25, 0.0625], 'offsets': [0, 1, 2]}
this = kernel_binom(5, 0, 0)
self._compare(res, this)
def test_3x3(self):
res = {'weights': [0.5, 0.25], 'offsets': [0, 1]}
this = kernel_binom(3, 0, 0)
self._compare(res, this)
def main():
parser = argparse.ArgumentParser(
description="Calculates gaussian kernel weights and offsets from a "
"binomial distribution and optionally adjust the weights "
"and offsets for a linearly-sampled gaussian blur shader.")
parser.add_argument(
"taps", default=5, type=int,
help="Specify the number of taps (kernel size)"
)
parser.add_argument(
"--expand", default=0, type=int,
help="How much to expand the tap count (expand outermost "
"coefficients)"
)
parser.add_argument(
"--reduce", default=0, type=int,
help="How many taps to discard at borders (eliminate outermost "
"coefficients)"
)
parser.add_argument(
"--linear", default=False, action='store_true',
help="Uses linear sampling to compute weights and offsets"
)
args = parser.parse_args()
taps = args.taps
exp = args.expand
red = args.reduce
linear = args.linear
print "Computing a", \
bold("{0}-tap".format(taps)), \
"filter kernel (+%(exp)s/-%(red)s) %(desc)s" % \
{
'taps': taps,
'exp': exp * 2,
'red': red * 2,
'desc': " (+linear reduction)" if linear else ""
}
print "Initial gaussian distribution: {0}".format(
bold(str(binom(taps - 1 + exp * 2)))
)
ntap = taps
res_discrete = kernel_binom(taps, exp, red)
res_linear = None
if linear:
res_linear = kernel_binom_linear(res_discrete)
ntap = (len(res_linear['weights']) * 2) - 1
if res_discrete is not None:
float_format = "{:.8f}"
print "Initial ", \
bold("{0}-tap".format(taps)), \
"filter kernel coefficients:".format(taps)
print "\tweights:", \
[float_format.format(x) for x in res_discrete["weights"]]
print "\toffsets:", \
[float_format.format(x) for x in res_discrete["offsets"]]
if linear and res_linear is not None:
print "\nOptimized", \
bold("{0}-tap".format(ntap)), \
"filter kernel coefficients:".format(ntap)
print "\tweights:", \
[float_format.format(x) for x in res_linear["weights"]]
print "\toffsets:", \
[float_format.format(x) for x in res_linear["offsets"]]