# Pyctools - a picture processing algorithm development kit.
# http://github.com/jim-easterbrook/pyctools
# Copyright (C) 2019-20 Pyctools contributors
#
# This program is free software: you can redistribute it and/or
# modify it under the terms of the GNU General Public License as
# published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
# General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see
# <http://www.gnu.org/licenses/>.
__all__ = ['UnsharpMask']
__docformat__ = 'restructuredtext en'
import cv2
import numpy
from pyctools.components.interp.gaussianfilter import GaussianFilter
from pyctools.components.interp.resizecore import resize_frame
from pyctools.core.config import ConfigBool, ConfigFloat
from pyctools.core.base import Transformer
from pyctools.core.types import pt_float
[docs]
class UnsharpMask(Transformer):
"""Enhance image detail using an unsharp mask.
The `unsharp mask`_ is computed by subtracting a `Gaussian blurred`_
image from the original image. Low amplitude detail can be removed
before the mask is added back to the image to sharpen it. This can
reduce the increase in noise when lots of sharpening is applied.
The ``amount`` parameter specifies how much sharpening to apply. It
is a real number rather than the percentage used in some software.
The ``radius`` parameter sets the standard deviation of the Gaussian
blurring filter.
To avoid discontinuities in the mask the ``threshold`` is used in a
"coring" function. Detail lower than the threshold is ignored,
detail above the threshold is reduced by the threshold value.
Another option to reduce noise is ``denoise``. This uses a 5x5
median filter as part of the mask computation.
Note that this component can also be used to soften an image using a
Gausssian filter. Set ``amount = -1``, ``threshold = 0``, and
``denoise = False``.
============= ===== ====
Config
============= ===== ====
``amount`` float Amount of sharpening to apply.
``radius`` float Size of blurring function.
``threshold`` float Don't sharpen low amplitude detail.
``denoise`` bool Median filter the detail to suppress noise.
============= ===== ====
.. _Gaussian blurred: https://en.wikipedia.org/wiki/Gaussian_blur
.. _unsharp mask: https://en.wikipedia.org/wiki/Unsharp_masking
"""
def initialise(self):
self.config['amount'] = ConfigFloat(value=1.0, decimals=2)
self.config['radius'] = ConfigFloat(value=2.0, decimals=1)
self.config['threshold'] = ConfigFloat(value=0.0, decimals=1)
self.config['denoise'] = ConfigBool()
def transform(self, in_frame, out_frame):
self.update_config()
amount = self.config['amount']
radius = self.config['radius']
threshold = self.config['threshold']
denoise = self.config['denoise']
data = in_frame.as_numpy(dtype=pt_float)
# median filter image before computing mask
if denoise:
mask = cv2.medianBlur(data, 5)
else:
mask = data
# blur data with Gaussian and subtract to make mask
h_filter = GaussianFilter.core(x_sigma=radius).as_numpy(dtype=pt_float)
v_filter = GaussianFilter.core(y_sigma=radius).as_numpy(dtype=pt_float)
mask = mask - resize_frame(resize_frame(
mask, h_filter, 1, 1, 1, 1), v_filter, 1, 1, 1, 1)
# core out mask values below threshold
if threshold > 0.0:
mask_p = mask - pt_float(threshold)
mask_p *= mask_p > pt_float(0)
mask_n = mask + pt_float(threshold)
mask_n *= mask_n < pt_float(0)
mask = mask_p + mask_n
# add some mask back to image
out_frame.data = data + (mask * amount)
# add audit
out_frame.set_audit(
self, 'data = UnsharpMask(data)\n', with_config=self.config)
return True