# Copyright (C) 2021 ISIS Rutherford Appleton Laboratory UKRI
# SPDX - License - Identifier: GPL-3.0-or-later
from __future__ import annotations
import traceback
import uuid
from enum import Enum, auto
from functools import partial
from itertools import groupby
from logging import getLogger
from time import sleep
from typing import TYPE_CHECKING
from uuid import UUID
import numpy as np
from PyQt5.QtWidgets import QApplication, QLineEdit
from mantidimaging.core.data import ImageStack
from mantidimaging.core.operation_history.const import OPERATION_HISTORY, OPERATION_DISPLAY_NAME
from mantidimaging.gui.mvp_base import BasePresenter
from mantidimaging.gui.utility import BlockQtSignals
from mantidimaging.gui.utility.common import operation_in_progress
from mantidimaging.gui.windows.stack_choice.presenter import StackChoicePresenter
from mantidimaging.gui.widgets.dataset_selector import DatasetSelectorWidgetView
from .model import FiltersWindowModel
APPLY_TO_180_MSG = "Operations applied to the sample are also automatically applied to the " \
"180 degree projection. Please avoid applying an operation unless you're" \
" absolutely certain you need to.\nAre you sure you want to apply to 180" \
" degree projection?"
FLAT_FIELDING = "Flat-fielding"
FLAT_FIELD_REGION = [-0.2, 1.2]
CROP_COORDINATES = "Crop Coordinates"
ROI_NORMALISATION = "ROI Normalisation"
if TYPE_CHECKING:
from mantidimaging.gui.windows.main import MainWindowView # pragma: no cover
from mantidimaging.gui.windows.operations import FiltersWindowView # pragma: no cover
from mantidimaging.gui.widgets.mi_mini_image_view.view import MIMiniImageView
REPEAT_FLAT_FIELDING_MSG = "Do you want to run flat-fielding again? This could cause you to lose data."
[docs]
class Notification(Enum):
REGISTER_ACTIVE_FILTER = auto()
APPLY_FILTER = auto()
APPLY_FILTER_TO_ALL = auto()
UPDATE_PREVIEWS = auto()
SCROLL_PREVIEW_UP = auto()
SCROLL_PREVIEW_DOWN = auto()
def _find_nan_change(before_image, filtered_image_data):
"""
Returns a bool array of all the placed where a NaN has been removed.
:param before_image: The before image.
:param filtered_image_data: The filter result.
:return: The indexes of the pixels with removed NaNs.
"""
before_nan = np.isnan(before_image)
after_nan = np.isnan(filtered_image_data)
nan_change = np.logical_and(before_nan, ~after_nan)
return nan_change
[docs]
def sub(x: tuple[int, int]) -> int:
"""
Subtracts two tuples. Helper method for generating the negative slice list.
:param x: The int tuple.
:return: The first value subtracted from the second value.
"""
return x[1] - x[0]
def _group_consecutive_values(slices: list[int]) -> list[str]:
"""
Creates a list of slices with negative indices in a readable format.
:param slices: The list of indices of the slices that contain negative values.
:return: A list of strings for the index ranges e.g. 1-5, 7-20, etc
"""
ranges = []
for _, iterable in groupby(enumerate(slices), sub):
rng = list(iterable)
if len(rng) == 1:
s = str(rng[0][1])
else:
s = f"{rng[0][1]}-{rng[-1][1]}"
ranges.append(s)
return ranges
[docs]
class FiltersWindowPresenter(BasePresenter):
view: FiltersWindowView
stack: ImageStack | None = None
divider = "------------------------------------"
def __init__(self, view: FiltersWindowView, main_window: MainWindowView):
super().__init__(view)
self.model = FiltersWindowModel(self)
self._main_window = main_window
self.original_images_stack: dict[UUID, ImageStack] = {}
self.applying_to_all = False
self.filter_is_running = False
self.prev_apply_single_state = True
self.prev_apply_all_state = True
@property
def main_window(self) -> MainWindowView:
return self._main_window
[docs]
def notify(self, signal):
try:
if signal == Notification.REGISTER_ACTIVE_FILTER:
self.do_register_active_filter()
elif signal == Notification.APPLY_FILTER:
self.do_apply_filter()
elif signal == Notification.APPLY_FILTER_TO_ALL:
self.do_apply_filter_to_dataset()
elif signal == Notification.UPDATE_PREVIEWS:
self.do_update_previews()
elif signal == Notification.SCROLL_PREVIEW_UP:
self.do_scroll_preview(1)
elif signal == Notification.SCROLL_PREVIEW_DOWN:
self.do_scroll_preview(-1)
except Exception as e:
self.show_error(e, traceback.format_exc())
getLogger(__name__).exception("Notification handler failed")
@property
def max_preview_image_idx(self):
if self.stack is None:
return 0
if not self.model.selected_filter.operate_on_sinograms:
return self.stack.num_images - 1
else:
return self.stack.num_sinograms - 1
[docs]
def set_stack_uuid(self, uuid: uuid.UUID | None):
if uuid is not None:
self.set_stack(self.main_window.get_stack(uuid))
else:
self.set_stack(None)
[docs]
def set_stack(self, stack: ImageStack | None):
self.stack = stack
# Update the preview image index
with BlockQtSignals([self.view]):
self.set_preview_image_index(0)
max_slice = self.max_preview_image_idx
self.view.previewImageIndex.setMaximum(max_slice)
self.view.previews.z_slider.set_range(0, max_slice)
for row_id in range(self.view.filterPropertiesLayout.count()):
widget = self.view.filterPropertiesLayout.itemAt(row_id).widget()
if isinstance(widget, DatasetSelectorWidgetView):
widget._handle_loaded_datasets_changed()
self.do_update_previews()
[docs]
def set_preview_image_index(self, image_idx: int):
"""
Sets the current preview image index.
"""
self.model.preview_image_idx = image_idx
# Set preview index spin box to new index
preview_idx_spin = self.view.previewImageIndex
with BlockQtSignals([preview_idx_spin]):
preview_idx_spin.setValue(self.model.preview_image_idx)
with BlockQtSignals([self.view.previews.z_slider]):
self.view.previews.z_slider.set_value(self.model.preview_image_idx)
# Trigger preview updating
self.view.auto_update_triggered.emit()
[docs]
def do_register_active_filter(self):
filter_name = self.view.get_selected_filter()
# Get registration function for new filter
register_func = self.model.filter_registration_func(filter_name)
# Register new filter (adding it's property widgets to the properties layout)
filter_widget_kwargs = register_func(self.view.filterPropertiesLayout, self.view.auto_update_triggered.emit,
self.view)
if filter_name == CROP_COORDINATES or filter_name == ROI_NORMALISATION:
self.init_roi_field(filter_widget_kwargs["roi_field"])
self.model.setup_filter(filter_name, filter_widget_kwargs)
self.view.clear_notification_dialog()
self.view.previews.link_before_after_histogram_scales(self.model.link_histograms())
if self.model.show_negative_overlay():
self.view.previews.add_negative_overlay()
else:
self.view.previews.hide_negative_overlay()
# Update the preview image index
with BlockQtSignals([self.view]):
max_slice = self.max_preview_image_idx
self.view.previewImageIndex.setMaximum(max_slice)
self.view.previews.z_slider.set_range(0, max_slice)
if self.model.preview_image_idx > max_slice:
self.set_preview_image_index(0)
[docs]
def filter_uses_parameter(self, parameter):
return parameter in self.model.params_needed_from_stack.values() if \
self.model.params_needed_from_stack is not None else False
[docs]
def do_apply_filter(self):
if self._already_run_flat_fielding():
if not self.view.ask_confirmation(REPEAT_FLAT_FIELDING_MSG):
return
if self.view.safeApply.isChecked():
with operation_in_progress("Safe Apply: Copying Data", "-------------------------------------", self.view):
self.original_images_stack = {self.stack.id: self.stack.copy()}
# if is a 180degree stack and a user says no, cancel apply filter.
if self.is_a_proj180deg(self.stack) and not self.view.ask_confirmation(APPLY_TO_180_MSG):
return
apply_to = [self.stack]
self._do_apply_filter(apply_to)
[docs]
def do_apply_filter_to_dataset(self):
if not self.view.ask_confirmation(
"Apply this filter to the dataset of the SELECTED stack in the Operations window?"):
return
selected_stack = self.stack
if selected_stack is None:
self.view.show_error_dialog("No stack is currently selected in the operations window.")
return
dataset_id = self.main_window.get_dataset_id_from_stack_uuid(selected_stack.id)
if dataset_id is None:
self.view.show_error_dialog("Could not find a dataset for the selected stack.")
return
current_dataset = self.main_window.get_dataset(dataset_id)
if current_dataset is None:
self.view.show_error_dialog(f"Dataset {dataset_id} not found.")
return
stacks = current_dataset.all
if not stacks:
self.view.show_error_dialog("No stacks found in that dataset.")
return
if self.view.safeApply.isChecked():
from mantidimaging.gui.utility.common import operation_in_progress
with operation_in_progress("Safe Apply: Copying Data", self.divider, self.view):
self.original_images_stack = {stk.id: stk.copy() for stk in stacks}
self.applying_to_all = True
self._do_apply_filter(stacks)
def _wait_for_stack_choice(self, new_stack: ImageStack, stack_uuid: UUID):
stack_choice = StackChoicePresenter(self.original_images_stack[stack_uuid], new_stack, self)
del self.original_images_stack[stack_uuid]
if self.model.show_negative_overlay():
stack_choice.enable_nonpositive_check()
stack_choice.show()
while not stack_choice.done:
QApplication.processEvents()
QApplication.sendPostedEvents()
sleep(0.05)
return stack_choice.use_new_data
[docs]
def is_a_proj180deg(self, stack_to_check: ImageStack):
return any(stack_to_check is stack for stack in self.main_window.get_all_180_projections())
def _post_filter(self, updated_stacks: list[ImageStack], task):
try:
use_new_data = True
negative_stacks = []
for stack in updated_stacks:
# Ensure there is no error if we are to continue with safe apply and 180 degree.
if task.error is None:
# otherwise check with user which one to keep
if self.view.safeApply.isChecked():
use_new_data = self._wait_for_stack_choice(stack, stack.id)
# if the stack that was kept happened to have a proj180 stack - then apply the filter to that too
if stack.has_proj180deg() and use_new_data and not self.applying_to_all:
if self.model.selected_filter.allow_for_180_projection:
# Apply to proj180 synchronously - this function is already running async
# and running another async instance causes a race condition in the parallel module
# where the shared data can be removed in the middle of the operation of another operation
self._do_apply_filter_sync([stack.proj180deg])
if np.any(stack.data < 0):
negative_stacks.append(stack)
self.applying_to_all = False
if task.error:
# task failed, show why
self.view.show_error_dialog(f"Operation failed: {task.error}")
elif use_new_data:
# Feedback to user
self.view.clear_notification_dialog()
self.view.show_operation_completed(self.model.selected_filter.filter_name)
selected_filter = self.view.get_selected_filter()
if selected_filter == CROP_COORDINATES:
self.init_roi_field(self.model.filter_widget_kwargs["roi_field"])
elif selected_filter == FLAT_FIELDING and negative_stacks:
self._show_negative_values_error(negative_stacks)
self.sync_histogram_levels()
else:
self.view.clear_notification_dialog()
self.view.show_operation_cancelled(self.model.selected_filter.filter_name)
finally:
self.view.filter_applied.emit()
self._set_apply_buttons_enabled(self.prev_apply_single_state, self.prev_apply_all_state)
self.filter_is_running = False
self.do_update_previews()
[docs]
def sync_histogram_levels(self):
p_low, p_high = self.view.preview_image_after.histogram.getHistogramRange()
self.view.preview_image_before.histogram.setHistogramRange(p_low, p_high)
self.view.preview_image_before.histogram.setLevels(p_low, p_high)
def _do_apply_filter(self, apply_to: list[ImageStack]):
self.filter_is_running = True
# Record the previous button states
self.prev_apply_single_state = self.view.applyButton.isEnabled()
self.prev_apply_all_state = self.view.applyToAllButton.isEnabled()
# Disable the apply buttons
self._set_apply_buttons_enabled(False, False)
self.model.do_apply_filter(apply_to, partial(self._post_filter, apply_to))
def _do_apply_filter_sync(self, apply_to):
self.model.do_apply_filter_sync(apply_to, partial(self._post_filter, apply_to))
[docs]
def do_update_previews(self) -> None:
if self.stack is None:
self.view.clear_previews()
return
is_new_data = self.view.preview_image_before.image_data is None
is_flat_fielding = self._flat_fielding_is_selected()
self.view.clear_previews(clear_before=False)
# Only apply the lock scale after image data has been set for the first time otherwise no region is shown
lock_scale = self.view.lockScaleCheckBox.isChecked() and not is_new_data
if lock_scale:
self.view.previews.record_histogram_regions()
if is_flat_fielding:
self.view.previews.after_region = FLAT_FIELD_REGION
if not self.model.selected_filter.operate_on_sinograms:
subset = self.stack.slice_as_image_stack(self.model.preview_image_idx)
squeeze_axis = 0
else:
if self.stack.num_projections < 2:
self.show_error("This filter requires a stack with multiple projections", "")
self.view.clear_previews()
return
subset = self.stack.sino_as_image_stack(self.model.preview_image_idx)
squeeze_axis = 1
# Take copies for display to prevent issues when the shared memory is cleaned
before_image = np.copy(subset.data.squeeze(squeeze_axis))
try:
if self.model.filter_widget_kwargs:
self.model.apply_to_images(subset)
except Exception as e:
msg = f"Error applying filter for preview: {e}"
self.show_error(msg, traceback.format_exc())
# Can't continue be need the before image drawn
self._update_preview_image(before_image, self.view.preview_image_before)
return
# Update image after first in order to prevent wrong histogram ranges being shared
filtered_image_data = np.copy(subset.data.squeeze(squeeze_axis))
if np.any(filtered_image_data < 0):
self._show_preview_negative_values_error(self.model.preview_image_idx)
self._update_preview_image(filtered_image_data, self.view.preview_image_after)
# Update image before
self._update_preview_image(before_image, self.view.preview_image_before)
self.view.previews.update_histogram_data()
if filtered_image_data.shape == before_image.shape:
diff = np.subtract(filtered_image_data, before_image)
if self.view.overlayDifference.isChecked():
nan_change = _find_nan_change(before_image, filtered_image_data)
self.view.previews.add_difference_overlay(diff, nan_change)
else:
self.view.previews.hide_difference_overlay()
if self.view.invertDifference.isChecked():
diff = np.negative(diff, out=diff)
self._update_preview_image(diff, self.view.preview_image_difference)
# Ensure all of it is visible if the lock zoom isn't checked
if not self.view.lockZoomCheckBox.isChecked() or is_new_data:
self.view.previews.auto_range()
if lock_scale:
self.view.previews.restore_histogram_regions()
if is_flat_fielding:
self.view.preview_image_after.histogram.setHistogramRange(mn=FLAT_FIELD_REGION[0],
mx=FLAT_FIELD_REGION[1])
# We must auto-range the before histogram as when the before and after histograms are linked the
# change to the after scale is also applied to the before scale
self.view.preview_image_before.histogram.autoHistogramRange()
else:
self.view.previews.autorange_histograms()
self.view.previews.set_histogram_log_scale()
@staticmethod
def _update_preview_image(image_data: np.ndarray, image: MIMiniImageView):
image.clear()
image.setImage(image_data)
[docs]
def get_filter_module_name(self, filter_idx):
return self.model.get_filter_module_name(filter_idx)
[docs]
def set_filter_by_name(self, filter_menu_name: str):
self.view.filterSelector.setCurrentText(filter_menu_name)
def _already_run_flat_fielding(self):
"""
:return: True if this is not the first time flat-fielding is being run, False otherwise.
"""
if not self._flat_fielding_is_selected():
return False
if OPERATION_HISTORY not in self.stack.metadata:
return False
return any(operation[OPERATION_DISPLAY_NAME] == FLAT_FIELDING
for operation in self.stack.metadata[OPERATION_HISTORY])
def _set_apply_buttons_enabled(self, apply_single_enabled: bool, apply_all_enabled: bool):
"""
Changes the state of the apply buttons before/after an operation is run.
:param apply_single_enabled: The desired state for the apply button.
:param apply_all_enabled: The desired state for the apply to all button.
"""
self.view.applyButton.setEnabled(apply_single_enabled)
self.view.applyToAllButton.setEnabled(apply_all_enabled)
[docs]
def init_roi_field(self, roi_field: QLineEdit):
"""
Sets the initial value of the crop coordinates line edit widget so that it doesn't contain values that are
larger than the image.
:param roi_field: The ROI field line edit widget.
"""
if self.stack is None:
return
x = self.stack.data.shape[1] // 2
y = self.stack.data.shape[2] // 2
crop_string = ", ".join(["0", "0", str(y), str(x)])
roi_field.setText(crop_string)
def _show_negative_values_error(self, negative_stacks: list[ImageStack]):
"""
Shows information on the view and in the log about negative values in the output.
:param negative_stacks: A list of stacks with negative values in the data.
"""
operation_name = self.model.selected_filter.filter_name
gui_error = [f"{operation_name} completed."]
for stack in negative_stacks:
negative_slices = []
for i in range(len(stack.data)):
if np.any(stack.data[i] < 0):
negative_slices.append(i)
stack_msg = f'Slices containing negative values in {stack.name}: '
if len(negative_slices) == len(stack.data):
slices_msg = stack_msg + "all slices."
gui_error.append(slices_msg)
else:
ranges = _group_consecutive_values(negative_slices)
slices_msg = stack_msg + f'{", ".join(ranges)}.'
if len(negative_slices) <= 12:
gui_error.append(slices_msg)
else:
gui_error.append(f'{stack_msg}{", ".join(ranges[:10])} ... {ranges[-1]}.')
getLogger(__name__).error(slices_msg)
self.view.show_error_dialog(" ".join(gui_error))
def _show_preview_negative_values_error(self, slice_idx: int):
"""
Shows a message that the operation preview contains negative values.
:param slice_idx: The index of the preview slice.
"""
self.view.show_error_dialog(f"Negative values found in result preview for slice {slice_idx}.")
def _flat_fielding_is_selected(self):
return self.view.get_selected_filter() == FLAT_FIELDING