def update(self, changed, thread):
wavelet = self.wavelet.get()
if wavelets.TriangeWaveletFamily().is_from(wavelet) or wavelets.BSplineWaveletFamily().is_from(wavelet):
dec = wtutils.uiwt
rec = wtutils.uiwt_inv
else:
dec = wtutils.uwt
rec = wtutils.uwt_inv
denoise = Denoise(wavelet, self.scale.get(), self.boundary, dec, rec, mode=self.mode.get(), thread=thread)
img = self.img.get()
if changed in [self.img, self.noise_level] or self.noisy is None:
self.noisy = img.data + nputils.gaussian_noise(img.data.shape, 0, self.noise_level.get())
estimated_noise_sigma = nputils.k_sigma_noise_estimation(self.noisy)
print "Estimated noise:", estimated_noise_sigma
denoised = denoise.do(self.noisy, estimated_noise_sigma, threashold_factor=self.threashold_factor.get())
# denoised = nputils.smooth(self.noisy, 9, mode="same", boundary="symm")
if not thread.is_alive() or denoised == None:
return False
return (img, imgutils.Image.from_image(img, self.noisy), imgutils.Image.from_image(img, denoised))
def get_bg(self, img):
"""Return either the noise level or a background map.
"""
if self.config.data.bg_use_ksigma_method:
return nputils.k_sigma_noise_estimation(img.data)
if self.config.data.bg_coords is not None:
x1, y1, x2, y2 = self.config.data.bg_coords
prj = self.get_projection(img)
xy_p1, xy_p2 = np.round(prj.s2p([(x1, y1), (x2, y2)])).astype(int)
ex = [0, img.data.shape[1]]
ey = [0, img.data.shape[0]]
xlim1, xlim2 = sorted([nputils.clamp(xy_p1[0], *ex), nputils.clamp(xy_p2[0], *ex)])
ylim1, ylim2 = sorted([nputils.clamp(xy_p1[1], *ey), nputils.clamp(xy_p2[1], *ey)])
return img.data[ylim1:ylim2, xlim1:xlim2].copy()
elif self.config.data.bg_fct is not None:
return self.config.data.bg_fct(self, img)
raise Exception("Bg extraction method need to be set")
def get_bg(self, img):
"""Return either the noise level or a background map.
"""
if self.config.data.bg_use_ksigma_method:
return nputils.k_sigma_noise_estimation(img.data)
if self.config.data.bg_coords is not None:
x1, y1, x2, y2 = self.config.data.bg_coords
prj = self.get_projection(img)
xy_p1, xy_p2 = np.round(prj.s2p([(x1, y1), (x2, y2)])).astype(int)
ex = [0, img.data.shape[1]]
ey = [0, img.data.shape[0]]
xlim1, xlim2 = sorted(
[nputils.clamp(xy_p1[0], *ex),
nputils.clamp(xy_p2[0], *ex)])
ylim1, ylim2 = sorted(
[nputils.clamp(xy_p1[1], *ey),
nputils.clamp(xy_p2[1], *ey)])
return img.data[ylim1:ylim2, xlim1:xlim2].copy()
elif self.config.data.bg_fct is not None:
return self.config.data.bg_fct(self, img)
raise Exception("Bg extraction method need to be set")
def do(self, img, noise_sigma=None, noise=None, threashold_factor=4):
print "Denoising..."
if noise is None and noise_sigma is None:
noise_sigma = nputils.k_sigma_noise_estimation(img)
res = self.decompose(img)
if res is None:
return None
for (i, frame) in enumerate(itertools.chain(*res[:-1])):
noise_factor = self.get_noise_factor(i, noise_sigma, noise)
if noise_factor == None:
return None
threashold = threashold_factor * noise_factor
mask = (abs(frame) < threashold)
frame[mask] = 0
if self.mode == "soft":
frame[~mask] = frame[~mask] - threashold
denoised = self.recompose(res, img)
print "Done"
return denoised
def do(self, img, noise_sigma=None, noise=None, threashold_factor=4):
print "Denoising..."
if noise is None and noise_sigma is None:
noise_sigma = nputils.k_sigma_noise_estimation(img)
res = self.decompose(img)
if res is None:
return None
for (i, frame) in enumerate(itertools.chain(*res[:-1])):
noise_factor = self.get_noise_factor(i, noise_sigma, noise)
if noise_factor is None:
return None
threashold = threashold_factor * noise_factor
mask = (abs(frame) < threashold)
frame[mask] = 0
if self.mode == "soft":
frame[~mask] = frame[~mask] - threashold
denoised = self.recompose(res, img)
print "Done"
return denoised
def update(self, changed, thread):
wavelet = self.wavelet.get()
if wavelets.TriangeWaveletFamily().is_from(
wavelet) or wavelets.BSplineWaveletFamily().is_from(wavelet):
dec = wtutils.uiwt
rec = wtutils.uiwt_inv
else:
dec = wtutils.uwt
rec = wtutils.uwt_inv
denoise = Denoise(wavelet,
self.scale.get(),
self.boundary,
dec,
rec,
mode=self.mode.get(),
thread=thread)
img = self.img.get()
if changed in [self.img, self.noise_level] or self.noisy is None:
self.noisy = img.data + nputils.gaussian_noise(
img.data.shape, 0, self.noise_level.get())
estimated_noise_sigma = nputils.k_sigma_noise_estimation(self.noisy)
print "Estimated noise:", estimated_noise_sigma
denoised = denoise.do(self.noisy,
estimated_noise_sigma,
threashold_factor=self.threashold_factor.get())
# denoised = nputils.smooth(self.noisy, 9, mode="same", boundary="symm")
if not thread.is_alive() or denoised is None:
return False
return (img, imgutils.Image.from_image(img, self.noisy),
imgutils.Image.from_image(img, denoised))
