batoolset 0.1.0__py3-none-any.whl

batoolset/GUI/paint2.py

@@ -0,0 +1,1345 @@
+# TODO maybe hide green pointer if mouse is out of the image --> dangerous
+# NB this will ultimately become the one and only paint widget in the app --> all others or most of them should disappear and be replaced by that
+# TODO need add the save but only for the mask edit item!!!
+import os
+
+from batoolset.pyqt.tools import getCtrlModifier
+from batoolset.settings.global_settings import set_UI # set the UI to qtpy
+set_UI()
+import traceback
+from qtpy.QtGui import QKeySequence
+from skimage.measure import label, regionprops
+from qtpy.QtCore import QRect, QTimer, Qt
+from qtpy.QtWidgets import QWidget, QApplication, QMessageBox
+from batoolset.dialogs.opensave import saveFileDialog
+from batoolset.draw.widgets.vectorial import VectorialDrawPane
+from qtpy.QtWidgets import QMenu
+from qtpy import QtCore, QtGui, QtWidgets
+from batoolset.img import toQimage, Img, get_white_bounds, RGB_to_int24, is_binary, auto_scale, save_as_tiff
+from batoolset.ta.measurements.measurements3D.get_point_on_surface_if_centroid_is_bad import point_on_surface
+from batoolset.ta.segmentation.neo_wshed import wshed
+from batoolset.ta.selections.selection import get_colors_drawn_over, convert_selection_color_to_coords
+from batoolset.files.tools import smart_name_parser
+import numpy as np
+from batoolset.tools.logger import TA_logger # logging
+
+logger = TA_logger()
+
+class Createpaintwidget(QWidget):
+    # auto_convert_float_to_binary_threshold = 0.4
+
+    default_mask_name = 'handCorrection.tif'
+
+    # maybe store
+    def __init__(self, enable_shortcuts=False): #, methods_overriding=None
+
+        super().__init__()
+        #
+        # if methods_overriding is not None:
+        #     for k,v in methods_overriding.items():
+        #         # replace internal method by external one
+        #         k = v
+        # self.image_path = None
+        self.save_path = None
+        self.vdp = VectorialDrawPane(active=False) # the vectorial drawing panel
+        self.raw_image = None # the original unprocessed image
+        self.image = None # the displayed image (a qimage representation of raw_image can be a subset of raw_image_also)
+        self.raw_mask = None  # the original unprocessed mask
+        self.imageDraw = None # the mask image drawn over the image
+        self.raw_user_drawing = None # contains just the user drawing without image mask
+        self.cursor = None
+        self.maskVisible = True
+        self.scale = 1.0
+        self.drawing = False
+        self.brushSize = 3
+        self.minimal_cell_size = 10
+        self._clear_size = 30
+        self.drawColor = QtGui.QColor(QtCore.Qt.red) # blue green cyan
+        self.eraseColor = QtGui.QColor(QtCore.Qt.black)
+        self.eraseColor_visualizer = QtGui.QColor(QtCore.Qt.green)
+        self.cursorColor = QtGui.QColor(QtCore.Qt.green)
+        self.lastPoint = QtCore.QPointF()
+        self.change = False
+        self.propagate_mouse_move = False
+        # KEEP IMPORTANT required to track mouse even when not clicked
+        self.setMouseTracking(True)  # KEEP IMPORTANT
+        self.scrollArea = None
+        self.statusBar = None
+        self.drawing_enabled = True
+        self.channel = None
+        self.force_cursor_visible = False # required to force activate cursor to be always visible
+        self.save_file_name=None
+        self.multichannel_mode = False
+
+        self.auto_convert_float_to_binary_threshold = 0.4
+
+        # self.multi_channel_editor_n_save = False
+        # TODO add sortcuts
+
+        if enable_shortcuts:
+            self.add_shortcuts() # TODO make this optional as it can cause trouble
+
+    def force_cursor_to_be_visible(self, boolean):
+        self.force_cursor_visible = boolean
+
+        # can be used for saving
+    def set_save_path(self, path):
+        self.save_path = path
+
+    def get_save_path(self):
+        return self.save_path
+
+    def set_scale(self, scale):
+        # print(self.scale) # 0.10000000000000014 --> this is still ok --> this value crashes 1.3877787807814457e-16
+
+        # make sure scale is bounded to avoid issues and non sense scalings
+        if scale<0.01:
+            scale = 0.01
+        if scale>50:
+            scale=50
+
+        self.scale = scale
+        self.vdp.scale =scale
+        # DO I NEED ADJUST MORE SCALES ???
+        # self.update()
+
+    def get_scale(self):
+        return self.scale
+
+    # comprend pas ça marche pas si parent
+    def add_shortcuts(self):
+        # zoomPlus = QtWidgets.QShortcut("Ctrl+Shift+=", self)
+        # zoomPlus.activated.connect(self.zoomIn)
+        # zoomPlus.setContext(QtCore.Qt.ApplicationShortcut)  # make sure the shorctut always remain active
+        #
+        # zoomPlus2 = QtWidgets.QShortcut("Ctrl++", self)
+        # zoomPlus2.activated.connect(self.zoomIn)
+        # zoomPlus2.setContext(QtCore.Qt.ApplicationShortcut)  # make sure the shorctut always remain active
+        #
+        # zoomMinus = QtWidgets.QShortcut("Ctrl+Shift+-", self)
+        # zoomMinus.activated.connect(self.zoomOut)
+        # zoomMinus.setContext(QtCore.Qt.ApplicationShortcut)  # make sure the shorctut always remain active
+        #
+        # zoomMinus2 = QtWidgets.QShortcut("Ctrl+-", self)
+        # zoomMinus2.activated.connect(self.zoomOut)
+        # zoomMinus2.setContext(QtCore.Qt.ApplicationShortcut)
+
+        # enter_pressed = QtWidgets.QShortcut("Ctrl+Shift+=", self)
+        # zoomPlus.activated.connect(self.zoomIn)
+        # zoomPlus.setContext(QtCore.Qt.ApplicationShortcut)  # make sure the shorctut always remain active
+
+
+        padEnterShortcut = QtWidgets.QShortcut(QtGui.QKeySequence(QtCore.Qt.Key_Enter), self)  # voila
+        padEnterShortcut.activated.connect(self.apply)
+        padEnterShortcut.setContext(QtCore.Qt.ApplicationShortcut)  # make s
+
+        enterShortcut = QtWidgets.QShortcut(QtGui.QKeySequence(QtCore.Qt.Key_Return), self)  # voila
+        enterShortcut.activated.connect(self.apply)
+        enterShortcut.setContext(QtCore.Qt.ApplicationShortcut)  # make s
+
+        enterShortcut2 = QtWidgets.QShortcut('Shift+Return', self)  # voila
+        enterShortcut2.activated.connect(self.apply)
+        enterShortcut2.setContext(QtCore.Qt.ApplicationShortcut)  # make s
+
+        # pb if parent has same shortcut then ignored
+        # the pb is really with enter --> maybe stolen focus by dialog ????
+        enterShortcut3 = QtWidgets.QShortcut('Shift+Enter', self)
+        enterShortcut3.activated.connect(self.apply)
+        enterShortcut3.setContext(QtCore.Qt.ApplicationShortcut)  # make sure the shorctut always remain active
+
+
+        self.shrtM = QtWidgets.QShortcut("M", self)
+        self.shrtM.activated.connect(self.m_apply)
+        self.shrtM.setContext(QtCore.Qt.ApplicationShortcut)
+
+
+        self.ctrl_shift_S_grab_screen_shot = QtWidgets.QShortcut('Ctrl+Shift+S', self)
+        self.ctrl_shift_S_grab_screen_shot.activated.connect(self.grab_screen_shot)
+        self.ctrl_shift_S_grab_screen_shot.setContext(QtCore.Qt.ApplicationShortcut)
+
+        self.increase_contrastC = QtWidgets.QShortcut('C', self)
+        self.increase_contrastC.activated.connect(self.increase_contrast)
+        self.increase_contrastC.setContext(QtCore.Qt.ApplicationShortcut)
+
+        # QtGui.QShortcut(
+        # self.supr = QtWidgets.QShortcut(QtGui.QKeySequence(QtCore.Qt.Key_Delete), self)
+
+        # QKeySequence(tr("Ctrl+X, Ctrl+C"))
+        # QKeySequence(Qt.CTRL + Qt.Key_X, Qt.CTRL + Qt.Key_C)
+
+        # self.supr = QtWidgets.QShortcut(QKeySequence(Qt.Key_Delete), self)
+        # self.supr.activated.connect(self.suppr_pressed)
+        # self.supr.setContext(QtCore.Qt.ApplicationShortcut)
+
+
+        # connect the m shortcut to show and hide mask
+
+    # this returns the coords of the cells drawn over --> very useful for clones, adding/removing cells such as dividing cells and dying cells !!!
+    # that seems to work very well --> TODO finalize that now or soon
+    def get_selection_coords_based_on_current_mask(self, bg_color=None):
+        coords_of_selection = []
+
+        # pb --> here I need do it on the labeled image -->
+
+        img_to_analyze = self.get_raw_image()
+        # si l'image est juste un mask binaire alors -->  faut la convertir en label --> TODO
+        if is_binary(img_to_analyze):
+            if len(img_to_analyze.shape)==3:
+                # assume it's a handCorrection mask and the mask is 3channel
+                img_to_analyze = img_to_analyze[...,0]
+            img_to_analyze = label(img_to_analyze, connectivity=1, background=255) # will that always work
+            bg_color = 0
+
+        # why is this shit empty
+        mask=self.get_user_drawing()
+
+        # --> seems to be ok --> then why 0
+        # import matplotlib.pyplot as plt
+        # plt.imshow(img_to_analyze)
+        # plt.show()
+
+        selected_cells =  get_colors_drawn_over(mask,  img_to_analyze)
+        if not selected_cells:
+            logger.warning('No selection found --> nothing to do')
+            return coords_of_selection
+
+        self.set_mask(np.zeros_like(mask))
+        self.update()
+
+        tmp = img_to_analyze
+        if len(tmp.shape) == 3:
+            tmp = RGB_to_int24(tmp)
+        #
+        if bg_color is None:
+            if 0xFFFFFF in tmp:
+                bg_color = 0xFFFFFF
+            else:
+                bg_color = 0
+        #
+        # cell_label = label(tmp, connectivity=1, background=bg_color)
+        #
+        # # for all the selected cells --> return the coords --> probably a quite good idea...
+        # for region in regionprops(cell_label):
+        #     color = tmp[region.coords[0][0], region.coords[0][1]]
+        #     if color in selected_cells:
+        #         point_awlays_inside_the_cell = point_on_surface(region, cell_label)
+        #         coords_of_selection.append(point_awlays_inside_the_cell)
+        #
+        # return coords_of_selection
+
+        return convert_selection_color_to_coords(tmp, selected_cells=selected_cells, bg_color=bg_color)
+
+    # added this so that I can easily get user selection using a cell ID
+    def get_colors_drawn_over(self, forbidden_colors=[0,0xFFFFFF]):
+        mask = self.get_user_drawing()
+        # print('I was called')
+        if mask is None:
+            return
+
+        selected_colors = get_colors_drawn_over(mask,  self.get_raw_image(),forbidden_colors=forbidden_colors)
+        # # print(mask)
+        # # need get all the colors of the raw image that correspond to user selection
+        # selected_colors = self.get_raw_image()[mask != 0]
+        # selected_colors = RGB_to_int24(selected_colors)
+        # selected_colors = set(selected_colors.ravel().tolist())
+        # # if 0xFFFFFF in selected_colors:
+        # #     selected_colors.remove(0xFFFFFF)
+        # if forbidden_colors is not None:
+        #     if not isinstance(forbidden_colors, list):
+        #         forbidden_colors = [forbidden_colors]
+        #     for color in forbidden_colors:
+        #         if color in selected_colors:
+        #             selected_colors.remove(color)
+        # selected_colors = list(selected_colors)
+        self.set_mask(np.zeros_like(mask))
+        self.update()
+        return selected_colors
+
+    # def suppr_pressed(self):
+
+    def increase_contrast(self):
+        # print('increasing contrast')
+        try:
+            # metadata = self.raw_image.metadata
+            # sqqsdq
+            # self.raw_image = Img(auto_threshold(self.raw_image),metadata=metadata)
+            # self.set_image(self.raw_image)
+            # shall I do it for the displayed image only ???? --> probably smarter porbably shoudl do a get display image that is independent of all other functions and handles all dims
+            # very dumb --> needs be done just for the dipslayed image and nothing else and --> maybe get the image back from the screen directly --> would be smarted and better
+
+            # maybe do this only on the displayed image --> using convert_qimage_to_numpy or do something that gets the image to be displayed including the channels
+            # or do a tool that will recover the image of interest with channels and also the rest of the dimensions --> MEGA TODO !!!
+            print('auto-increase contrast')
+            # hack to preserve luts --> TODO
+            copy = auto_scale(np.copy(self.raw_image))
+            meta = None
+            try:
+                meta=self.raw_image.metadata
+            except:
+                pass
+            # print(self.raw_image.metadata)
+            self.set_display(copy, metadata=meta)
+        except:
+            traceback.print_exc()
+
+    def grab_screen_shot(self):
+            # screenshot = QPixmap.grabWindow(self.winId())
+            output_file = saveFileDialog(parent_window=self, extensions="Supported Files (*.png);;All Files (*)",
+                                         default_ext='.png')
+            if output_file is not None:
+
+                try:
+                    # we hide the cursor
+                    self.cursor = QtGui.QImage(self.image.size(), QtGui.QImage.Format_ARGB32)
+                    self.cursor.fill(QtCore.Qt.transparent)
+                    self.update()
+                    screenshot=self.grab()
+                    screenshot.save(output_file, 'png')
+                    # restore cursor
+                    # or paint without cursor --> TODO do that better some day
+                except:
+                    logger.error('Could not grab a screenshot...')
+
+    def save(self):
+        # save_img
+        print('ctrl S save method to override')
+
+    def apply(self):
+        print('apply method to override')
+
+    def shift_apply(self):
+        print('shift apply method to override')
+
+    def ctrl_m_apply(self):
+        print('ctrl m apply method to override')
+
+    def suppr_pressed(self):
+        # print('Suppr method to override')
+        if self.vdp.active:
+            self.vdp.removeCurShape()
+            self.update()
+        else:
+            print('Suppr method to override')
+
+
+    # def m_apply(self):
+    #     print('m apply method to override')
+    #     qsdsqqsdsq
+    def m_apply(self):
+        # print('m apply called')
+        self.maskVisible = not self.maskVisible
+        self.update()
+
+    def set_display(self, display, metadata=None):
+        # this actually sets the displayed image --> different from set imlage, can be used to show channels instead or stuff like that instead of the actula imge
+        if isinstance(display, np.ndarray):
+            display = toQimage(display, metadata=metadata)
+        self.image = display
+        self.update()
+
+    def set_image(self, img):
+        if img is None:
+            self.save_file_name = None
+            self.raw_image = None
+            self.image = None
+            self.imageDraw = None
+            self.raw_mask = None
+            self.raw_user_drawing = None
+            self.unsetCursor()  # restore cursor
+            self.update()
+            return
+        else:
+                # remove cursor only if image is shown and not vectorial drawing mode!
+                if self.vdp.active:
+                    self.unsetCursor()
+                else:
+                    self.setCursor(Qt.BlankCursor)
+                # try:
+                #     if isinstance(img, np.ndarray):
+                #         self.image = img.getQimage() # bug is here
+                #         # set an empty mask
+                #         self.imageDraw = QtGui.QImage(self.image.size(), QtGui.QImage.Format_ARGB32)
+                #         self.imageDraw.fill(QtCore.Qt.transparent)
+                #     else:
+                #         self.image = img[0].getQimage()
+                #         # set the mask
+                #         self.imageDraw = Img(self.createRGBA(img[1].getQimage()), dimensions='hwc').getQimage()
+                # except:
+                # NB THIS IS VERY REDUNDNAT ---> STICK TO ONE METHOD
+                if isinstance(img, str):
+                    # self.image_path = img# maybe this image path can be used for saving, could also be provided serparately
+                    self.save_file_name = smart_name_parser(img, ordered_output=self.default_mask_name)
+                    img = Img(img)
+
+                # self.image_path=None # maybe this image path can be used for saving
+                if isinstance(img, np.ndarray):
+
+                    # bug is exactly here --> why and where
+
+                    # dimensions = None
+                    # if isinstance(img, Img):
+                    #     dimensions = img.get_dimensions_as_string()
+
+                    # self.image_path = None
+                    self.raw_image = img
+
+                    # print('tests',img.shape, isinstance(img,Img))
+
+                    self.image = toQimage(img)# maybe this image path can be used for saving
+                    if self.image is None:
+                        logger.error('Image could not be displayed...')
+                        self.unsetCursor()
+                        self.update()
+                        return
+                    self.imageDraw = QtGui.QImage(self.image.size(), QtGui.QImage.Format_ARGB32)
+                    self.imageDraw.fill(QtCore.Qt.transparent)
+                    self.raw_user_drawing =QtGui.QImage(self.image.size(), QtGui.QImage.Format_ARGB32)
+                    self.raw_user_drawing.fill(QtCore.Qt.transparent)
+                else:
+                    # self.image_path = None
+                    self.image = toQimage(img[0])
+                    self.raw_image = img[0]  #
+                    # same as set mask --> how to do that
+                    self.imageDraw = toQimage(Img(self.createRGBA(img[1]), dimensions='hwc'),preserve_alpha=True)#.getQimage()
+                    self.raw_user_drawing = QtGui.QImage(self.imageDraw.size(), QtGui.QImage.Format_ARGB32)
+                    self.raw_user_drawing.fill(QtCore.Qt.transparent)
+
+        if self.force_cursor_visible:
+            self.unsetCursor()
+
+        # self.image = QPixmap(100,200).toImage()
+        width = self.image.size().width()
+        height = self.image.size().height()
+        top = self.geometry().x()
+        left = self.geometry().y()
+        self.setGeometry(top, left, int(width*self.scale), int(height*self.scale))
+
+        self.cursor = QtGui.QImage(self.image.size(), QtGui.QImage.Format_ARGB32)
+        self.cursor.fill(QtCore.Qt.transparent)
+        self.update()
+
+    def get_raw_image(self):
+        return self.raw_image
+
+    def binarize(self, mask, auto_convert_float_to_binary=None, force=False):
+        if auto_convert_float_to_binary is None:
+            auto_convert_float_to_binary = self.auto_convert_float_to_binary_threshold
+        if auto_convert_float_to_binary and (mask.max() <= 1 or force):
+            # print('autoconvert')
+            mask = mask > auto_convert_float_to_binary
+        return mask
+
+    # do I need that ??? maybe yes if I just wanna change the mask
+    def set_mask(self, mask, auto_convert_float_to_binary=None):
+        if auto_convert_float_to_binary is None:
+            auto_convert_float_to_binary = self.auto_convert_float_to_binary_threshold
+        if isinstance(mask, str):
+            self.save_file_name = mask
+            # self.image_path = img# maybe this image path can be used for saving, could also be provided serparately
+            mask = Img(mask)
+            self.raw_mask = mask
+            # for TA compat
+            if mask.has_c():
+                # but should also save the raw mask as for the raw_image --> TODO --> maybe
+                if self.channel is not None:
+                    mask=mask[...,self.channel]
+                else:
+                    mask = mask[..., 0]
+
+        # convert mask to image draw
+        if mask is None:
+            self.imageDraw = None
+            self.raw_user_drawing = None
+            self.raw_mask = None
+        else:
+            # auto convert float deep learning masks to binary --> maybe this should be a parameter ...
+            # print('tada', auto_convert_float_to_binary, mask.max())
+
+            mask = self.binarize(mask, auto_convert_float_to_binary=auto_convert_float_to_binary)
+
+
+            # print('mask.shape',mask.shape)
+            # convert mask to a mask
+            # self.imageDraw = QtGui.QImage(self.image.size(), QtGui.QImage.Format_ARGB32)
+            # self.imageDraw.fill(QtCore.Qt.transparent)
+
+            # print('in here tmp ', mask.shape)
+
+
+            # if image has z channels and is more than just 2D / is 3D I could make this smarter by loading just the right image --> TODO --> sync the channel here with the channel of the currently dispalyer image
+            # SHALL I MAKE A FULL GUI FOR THAT ON RELY ON WHAT I HAVE ALREADY !!!!
+            self.imageDraw = toQimage(Img(self.createRGBA(mask), dimensions='hwc'),preserve_alpha=True) #.getQimage()  # marche pas car besoin d'une ARGB
+            self.raw_user_drawing = QtGui.QImage(self.imageDraw.size(), QtGui.QImage.Format_ARGB32)
+            self.raw_user_drawing.fill(QtCore.Qt.transparent) # somehow this is really required to have an empty image otherwise it is not --> ??? why
+        self.update()
+
+
+    # do I have a way to just get what was drawn by the user and the bounds of it --> maybe by substratction --> or can I get the difference
+    # I need bounds of drawn stuff --> think about how to do that !!!
+    # we have used the user drawing for something --> let's reset it
+    # shall I also save the raw mask unperturbed by user drawing --> can be useful too in fact to do things with it
+    def reset_user_drawing(self):
+        if self.imageDraw is None:
+            self.raw_user_drawing = None
+        else:
+            self.raw_user_drawing = QtGui.QImage(self.imageDraw.size(), QtGui.QImage.Format_ARGB32)
+            self.raw_user_drawing.fill(QtCore.Qt.transparent)
+
+            # shall I draw any user interaction also in a separate file so that I can get just what was drawn without having the original mask ??? --> maybe a good idea --> call it get user drawing ???
+    def get_mask(self):
+        # nb will only work for red pen !!!! --> do not allow to change pen color
+        # pas mal mais image est RGBA
+        if self.imageDraw is not None:
+            # return Img.qimageToNumpy(self.imageDraw, mode='bnw8')
+            mask = self.convert_qimage_to_numpy(self.imageDraw)[..., 2]
+            # for some reason there are some 1 in the image --> needs a fix
+            mask[mask!=255]=0
+            return mask
+        else:
+            return None
+
+    def get_user_drawing(self, show_erased=False):
+        if self.raw_user_drawing is not None:
+
+
+
+            # bug here the user draing contains the original mask --> bug
+
+            all = self.convert_qimage_to_numpy(self.raw_user_drawing)
+
+            # plt.imshow(all)
+            # plt.show()
+
+            # no clue why this stuff is empty
+            # import matplotlib.pyplot as plt
+            # plt.imshow(all)
+            # plt.show()
+
+            drawn = all[..., 2]
+
+
+            # plt.imshow(erased) #--> ok
+            # plt.show()
+
+            # plt.imshow(drawn)
+            # plt.show()
+
+
+            drawn[drawn != 255] = 0
+
+            if show_erased:
+                erased = all[..., 1]
+                erased[erased != 255] = 0
+                return [erased, drawn]
+
+            return drawn
+        else:
+            return None
+
+    # from https://stackoverflow.com/questions/19902183/qimage-to-numpy-array-using-pyside
+    # nb there seems to be a bug in the order of the channels
+    def convert_qimage_to_numpy(self, qimage):
+        qimage = qimage.convertToFormat(QtGui.QImage.Format.Format_RGB32)
+        width = qimage.width()
+        height = qimage.height()
+        image_pointer = qimage.bits() # creates a deep copy --> this is what I want
+        try:
+            image_pointer.setsize(qimage.sizeInBytes()) # qt6 version of the stuff
+        except:
+            image_pointer.setsize(qimage.byteCount())
+
+
+        # arr = np.array(image_pointer,copy=True).reshape(height, width, 4)
+        arr = np.array(image_pointer).reshape(height, width, 4)
+        return arr
+
+    def createRGBA(self, handCorrection):
+
+
+
+
+
+        # print('Im called')
+        # use pen color to display the mask
+        # in fact I need to put the real color
+        # nb this will definitely not work if the user wants to have full 3D stuff
+
+        # shall I also offer the zooming on the objects --> ????
+
+
+
+
+        RGBA = np.zeros((handCorrection.shape[0], handCorrection.shape[1], 4), dtype=np.uint8)
+
+        red = self.drawColor.red()
+        green = self.drawColor.green()
+        blue = self.drawColor.blue()
+
+        #(handCorrection.shape[0], handCorrection.shape[1], 4))
+        # mask = handCorrection[handCorrection==255]
+
+        # ce truc est bleu --> pkoi
+        # RGBA[..., 0] = 255#handCorrection --> totalement transp --> transparence color --> GREEN --> BLUE
+        # RGBA[..., 1] = 0  # this is the red channel --> RED --> GREEN
+        # RGBA[..., 2] = 0 # rien ici --> tt transparent --> alpha ??? --> ALPHA -->RED
+        # RGBA[..., 3] = 255 # --> BLUE --> ALPHA --> ok
+
+        # pkoi je swappe les channels avant
+
+        # 255 partout --> blanc
+        # 0 255 255 0 --> rouge --> 1 = red channel and 2 = alpha --> weird
+
+        # RGBA[RGBA[..., 0] > 0, 0] = blue
+        # RGBA[RGBA[..., 1] > 0, 1] = green
+        # RGBA[RGBA[..., 2] > 0, 2] = red
+        # RGBA[..., 3] = 255 # ne s'affiche pas et je comprend pas pkoi
+
+        #BGR in fact --> need fix here or the other --> the one in
+
+
+        # how can I do that in a simpler way ???
+        # bug somewhere in qimage --> fix it some day --> due to bgra instead of RGBA
+        # assumes image has just the first
+        RGBA[handCorrection != 0, 0] = blue # b
+        # RGBA[..., 1] = handCorrection
+        RGBA[handCorrection != 0, 1] = green # g
+        RGBA[handCorrection != 0, 2] = red # r
+        RGBA[..., 3] = 255 # alpha --> indeed alpha
+        RGBA[handCorrection == 0, 3] = 0 # ça marche maintenant --> super complexe qd meme je trouve
+
+        # can I do it in one go --> I guess yes!!!
+        # RGBA[..., 3] = 255
+        # RGBA[handCorrection != 0] = blue,green,red, 0
+        # RGBA[handCorrection==0, 3] = 255
+
+        # ok mais plus transparent
+
+        #
+        # print(red, green, blue)
+
+        # RGBA[RGBA[..., 0] == 255] = red
+        # RGBA[RGBA[..., 1] == 255] = green
+        # RGBA[RGBA[..., 2] == 255] = blue
+
+        return RGBA
+
+        # TODO also implement channel change directly within the display tool
+        # if merge is applied --> apply on average of all channels --> maybe not so smart an idea but ok to start with and better than what I do in TA
+
+    # dunno if I put it here or not ???
+    # def get_nb_channels(self):
+    #     if self.
+
+    # depreacted --> ideally should not be used
+    # TODO--> maybe implement a multi channel save -> TODO
+    # skip
+    def channelChange(self, i, skip_update_display=False):
+        # update displayed image depending on channel
+        # dqqsdqsdqsd
+        # pass
+        # try change channel if
+
+        # print('in channel change !!!')
+        # if image_to_display is None:
+        #     image_to_display = self.raw_image
+        if self.raw_image is not None:
+            # print('in', self.img.metadata)
+            # if self.Stack.currentIndex() == 0:
+                # need copy the image --> implement that
+                # print(self.img[..., i].copy())
+                # print(self.img[..., i])
+                if i == 0 and not skip_update_display:
+                    self.set_display(self.raw_image)
+                    self.channel = None
+                    # print('original', self.img.metadata)
+                else:
+                    if not skip_update_display:
+                        # print('modified0', self.img.metadata)
+                        # I need a hack when the image is single channel yet I need several masks for it !!!
+                        meta = None
+                        try:
+                            meta = self.raw_image
+                        except:
+                            pass
+                        if self.multichannel_mode and i-1>=self.raw_image.shape[-1]:
+                            channel_img = self.raw_image.imCopy(c=0) # if out of bonds load the first channel
+                        else:
+                            channel_img = self.raw_image.imCopy(c=i - 1)  # it's here that it is affected
+                        self.set_display(
+                                channel_img, metadata=meta)  # maybe do a set display instead rather --> easier to handle --> does a subest of the other
+                    self.channel = i-1
+                    # print('modified1', self.img.metadata)
+                    # print('modified2', channel_img.metadata)
+
+                    if self.multichannel_mode:
+                        if self.raw_mask is not None:
+                            # print('multichannel_mode', self.multichannel_mode)
+                            try:
+                                # print('self.raw_mask.shape',self.raw_mask.shape)
+                                if len(self.raw_mask.shape)<=2:
+                                    self.raw_mask = self.raw_mask[..., np.newaxis]
+                                self.set_mask(self.raw_mask[..., i-1])
+                            except:
+                                # print(i-1,  self.raw_mask.shape[-1] , len(self.raw_mask.shape))
+                                if i-1 >= self.raw_mask.shape[-1] and len(self.raw_mask.shape)>2:
+                                    # extend mask
+
+
+                                    if len(self.raw_mask.shape) > 2:
+                                        raw_mask2 = np.zeros_like(self.raw_mask, shape=(*self.raw_mask.shape[0:-1], i))
+                                    else:
+                                        raw_mask2 = np.zeros_like(self.raw_mask, shape=(*self.raw_mask.shape, i))
+
+                                    # copy existing mask to it
+                                    # print(raw_mask2.shape, self.raw_mask.shape)
+                                    for ch in range(self.raw_mask.shape[-1]):
+                                        raw_mask2[...,ch]=self.raw_mask[...,ch]
+                                    self.raw_mask = raw_mask2
+                                    self.set_mask(self.raw_mask[..., i - 1])
+                                else:
+                                    traceback.print_exc()
+                        else:
+                            # create empty mask
+                            if len(self.raw_image.shape)>2:
+                                self.raw_mask = np.zeros(shape=(*self.raw_image.shape[0:-1],i))
+                            else:
+                                self.raw_mask = np.zeros(shape=(*self.raw_image.shape, i))
+
+                            # print('doubs', self.raw_mask.shape, self.raw_image.shape)
+                            self.set_mask(self.raw_mask[..., i - 1])
+                self.update()
+        else:
+            self.channel = None
+            # else:
+            #     # logger.error("Not implemented yet TODO add support for channels in 3D viewer")
+            #     # sdqdqsdsqdqsd
+            #     self.loadVolume()
+
+
+    def get_colors_drawn_over(self):
+        mask = self.get_user_drawing()
+        # print('I was called')
+        if mask is None:
+            return []
+        # print(mask)
+
+        # need get all the colors of the raw image that correspond to user selection
+        selected_colors =self.get_raw_image()[mask!=0]
+        selected_colors = RGB_to_int24(selected_colors)
+
+        selected_colors =set(selected_colors.ravel().tolist())
+        if 0xFFFFFF in selected_colors:
+            selected_colors.remove(0xFFFFFF)
+        selected_colors = list(selected_colors)
+        # this is the only way I found to reset the user drawing but this is dirty and I must find a better  way
+        self.set_mask(np.zeros_like(mask))
+
+        self.update()
+        return selected_colors
+
+    def edit_drawing(self):
+        # if self.multichannel_mode and self.channel is None:
+        #     logger.error('please select a channel first')
+        #     return
+
+        # basic mask drawing edit
+        drawn_mask = self.get_mask()
+        if drawn_mask is None:
+            return
+        drawn_mask = np.copy(drawn_mask)
+        erased, drawn = self.get_user_drawing(show_erased=True)
+        # bounds = get_white_bounds([drawn, erased])
+        # if bounds is None:
+        # pass
+        # print(drawn_mask.shape)
+        # print(drawn_mask.max(), drawn_mask.min())
+        # print(drawn_mask.dtype)
+
+        drawn_mask[erased!=0]=0
+        drawn_mask[drawn!=0]=drawn[drawn!=0]
+        if self.multichannel_mode:
+            if self.raw_mask is not None:
+                if self.channel is not None:
+                    # print('overwriting mask in image')
+                    self.raw_mask[..., self.channel]=drawn_mask
+                    # big bug here this stuff contains the bg image --> really huge bug
+                    # Img(self.raw_mask, dimensions='hwc').save('/E/Sample_images/sample_images_PA/test_complete_wing_raphael/tst2.tif')
+                else:
+                    # if nothing is specified assume ch 0
+                    if len(self.raw_mask.shape)>2:
+                        self.raw_mask[..., 0] = drawn_mask
+                    else:
+                        self.raw_mask = drawn_mask
+
+        # Img(drawn_mask, dimensions='hw').save('/E/Sample_images/sample_images_PA/test_complete_wing_raphael/tst.tif')
+        self.set_mask(Img(drawn_mask, dimensions='hw'))
+
+
+    # draws/removes bonds locally à la TA
+    def apply_drawing(self, minimal_cell_size=0):
+        # print('minimal_cell_size', minimal_cell_size)
+
+        drawn_mask = self.get_mask()
+        # print(drawn_mask)
+        # no image or nothing drawn --> return
+        if drawn_mask is None:
+            return
+        erased, drawn = self.get_user_drawing(show_erased=True)
+        bounds = get_white_bounds([drawn, erased])
+        if bounds is None:
+            # in case the guy presses shift apply with no mask then run it over the whole image --> can make sense to have this
+            if minimal_cell_size is not None and minimal_cell_size !=0:
+                # dirty way of changing size --> needs be improved
+                minimal_cell_size = self.minimal_cell_size
+
+
+                # print('minimal_cell_size2', minimal_cell_size)
+                drawn_mask = wshed(drawn_mask, seeds='mask', min_seed_area=minimal_cell_size)
+                self.set_mask(drawn_mask)
+            return
+
+        if np.count_nonzero(erased) != 0:
+            lab_erased = label(erased, connectivity=2, background=0)
+            rps_lab_erased = regionprops(lab_erased)
+
+            lab_drawn_mask = label(drawn_mask, connectivity=1, background=255)
+            rps_lab_drawn_mask = regionprops(lab_drawn_mask)
+
+            for erased_bond in rps_lab_erased:
+                ids = np.unique(lab_drawn_mask[erased_bond.coords[:, 0], erased_bond.coords[:, 1]])
+                for id in ids:
+                    if id != 0:
+                        cur_bounds = rps_lab_drawn_mask[id - 1].bbox
+                        bounds[0] = min(bounds[0], cur_bounds[0])
+                        bounds[1] = max(bounds[1], cur_bounds[2])
+                        bounds[2] = min(bounds[2], cur_bounds[1])
+                        bounds[3] = max(bounds[3], cur_bounds[3])
+
+        min_y, max_y, min_x, max_x = bounds
+
+        first_x = 128
+        first_y = 128
+        min_x = min_x - 128
+        max_x = max_x + 128
+        min_y = min_y - 128
+        max_y = max_y + 128
+
+        # TODO fix upper bond too
+
+        if min_x < 0:
+            first_x = 128 + min_x
+            min_x = 0
+        if min_y < 0:
+            first_y = 128 + min_y
+            min_y = 0
+        if max_y > drawn_mask.shape[0]:
+            max_y = drawn_mask.shape[0]
+        if max_x > drawn_mask.shape[1]:
+            max_x = drawn_mask.shape[1]
+
+        minished = wshed(drawn_mask[min_y:max_y, min_x:max_x], seeds='mask', min_seed_area=minimal_cell_size)
+        drawn_mask[min_y:max_y, min_x:max_x][first_y:min(first_y + (bounds[1] - bounds[0]) + 1, minished.shape[0]),first_x:min(first_x + (bounds[3] - bounds[2]) + 1, minished.shape[1])] = minished[first_y:min(first_y + (bounds[1] - bounds[0]) + 1, minished.shape[0]), first_x:min(first_x + (bounds[3] - bounds[2]) + 1, minished.shape[1])]
+        self.set_mask(drawn_mask)
+
+    def force_consistent_range(self, img, auto_convert_float_to_binary=None):
+        if auto_convert_float_to_binary is None:
+            auto_convert_float_to_binary = self.auto_convert_float_to_binary_threshold
+        # for ch in range(img.shape[-1]):
+        #     tmp = img[...,ch]
+        #     min = tmp.min()
+        #     max = tmp.max()
+        #     # if min!=max:
+        #         # change range of the stuff
+        #         # print('error in range')
+        #     if max<=1.0:
+        #         print('error in range')
+        #     print(min, max)
+        #     print(min, max)
+        # img = 
+        return self.binarize(img, auto_convert_float_to_binary=auto_convert_float_to_binary, force=True)
+
+    def save_mask(self, multichannel_save=False, forced_nb_of_channels=None):
+        # print('saving to ....' + self.save_file_name)
+        if self.save_file_name is not None:
+            # print('really saving/...')
+            mask = self.get_mask()
+            if mask is not None:
+                logger.debug('saving mask to '+str(self.save_file_name))
+                try:
+                    single_channel_image = len(self.raw_image.shape)==2
+                except:
+                    single_channel_image = False
+                if multichannel_save == False or single_channel_image  and (forced_nb_of_channels is None or forced_nb_of_channels ==0):
+                    Img(mask, dimensions='hw').save(self.save_file_name)
+                else:
+                    if self.raw_mask is not None:
+                        # print('quick saving edited mask')
+                        # Img(self.raw_mask, dimensions='hwc').save(self.save_file_name)
+
+                        if len(self.raw_mask.shape)>2:
+                            # I need check all channels have the same normalization
+                            Img(self.force_consistent_range(self.raw_mask).astype(np.uint8)*255, dimensions='hwc').save(self.save_file_name)
+                        else:
+                            Img(self.raw_mask, dimensions='hw').save(self.save_file_name)
+                        return
+
+                    # print('test', self.channel )
+                    # print('orig img channels', self.raw_image.shape[-1])
+                    # if self.channel is None:
+                    #     logger.error('Please select a channel first')
+                    #     return
+                    # BELOW IS CRAPPY CODE --> BEST is if multichannel_mode to create a black empty image
+                    # if os.path.exists(self.save_file_name):
+                    #     try:
+                    #         mask_out = Img(self.save_file_name)
+                    #         if self.channel is not None:
+                    #             mask_out[...,self.channel]=mask
+                    #         else:
+                    #             # assume ch0
+                    #             mask_out[..., 0] = mask
+                    #         Img(mask_out, dimensions='hwc').save(self.save_file_name)
+                    #         return
+                    #     except:
+                    #         traceback.print_exc()
+                    #     print('le fichier existe')
+                    #
+                    #     # on le charge
+                    # # else:
+                    # print('le fichier existe pas')
+                    # mask_out = np.zeros_like(mask, shape=(*mask.shape, self.raw_image.shape[-1]))
+                    # if self.channel is not None:
+                    #     mask_out[..., self.channel] = mask
+                    # else:
+                    #     # assume ch0
+                    #     mask_out[..., 0] = mask
+                    # Img(mask_out, dimensions='hwc').save(self.save_file_name)
+                    # sinon on recup le nb de canaux de l'image finale
+
+
+    # def _update_channel(self, output, mask, channel):
+    #     if channel is not None:
+    #         mask_out = np.zeros_like(mask, shape=(*mask.shape, self.raw_image.shape[-1]))
+    #         mask_out[..., self.channel] = mask
+    #         Img(mask_out, dimensions='hwc').save(self.save_file_name)
+
+
+
+
+
+    # this is the local seeded watershed à la TA
+    def manually_reseeded_wshed(self):
+
+        if self.get_raw_image() is None:
+            return
+
+        try:
+            if self.channel is None and self.get_raw_image().has_c():
+                logger.error('Please select a channel fisrt')
+                return
+        except:
+            print(self.get_raw_image().shape, self.channel)
+            if self.channel is None and len(self.get_raw_image().shape)>=3 and self.get_raw_image().shape[-1]>1:
+                logger.error('Please select a channel fisrt')
+                return
+
+        usr_drawing = self.get_user_drawing()
+        labs = label(usr_drawing, connectivity=2, background=0)
+
+        # pb here because the user drawing also contains the mask and I don't want it
+        # plt.imshow(labs)
+        # plt.show()
+
+        rps_user_drawing = regionprops(labs)
+        if len(rps_user_drawing) == 0:
+            return
+
+        drawn_mask = self.get_mask()
+
+        # plt.imshow(drawn_mask)
+        # plt.show()
+
+        drawn_mask[labs != 0] = 0
+        lab_cells_user_drawing = label(drawn_mask, connectivity=1, background=255)
+        rps_cells_user_drawing = regionprops(lab_cells_user_drawing)
+
+        min_y = 100000000
+        min_x = 100000000
+        max_y = 0
+        max_x = 0
+        for rps_user in rps_user_drawing:
+            label_id = lab_cells_user_drawing[rps_user.coords[0][0], rps_user.coords[0][1]]
+            bbox = rps_cells_user_drawing[label_id - 1].bbox
+            min_y = min(bbox[0], min_y)
+            min_x = min(bbox[1], min_x)
+            max_y = max(bbox[2], max_y)
+            max_x = max(bbox[3], max_x)
+
+
+
+
+        # here I need get the channel in fact
+        img = self.get_raw_image()[min_y:max_y + 1, min_x:max_x + 1]
+        if self.channel is not None:
+            img = img[..., self.channel]
+        minished = wshed(img,
+                         seeds=labs[min_y:max_y + 1, min_x:max_x + 1], weak_blur=1.)
+        drawn_mask[min_y:max_y + 1, min_x:max_x + 1][minished != 0] = minished[minished != 0]
+        # plt.imshow(drawn_mask)
+        # plt.show()
+        self.set_mask(drawn_mask)
+
+
+    def mousePressEvent(self, event):
+        if not self.hasMouseTracking() or not self.drawing_enabled:
+            return
+        self.clickCount = 1
+        if self.vdp.active:
+            self.vdp.mousePressEvent(event)
+            self.update()
+            return
+
+        if event.buttons() == QtCore.Qt.LeftButton or event.buttons() == QtCore.Qt.RightButton:
+            self.drawing = True
+            zoom_corrected_pos = event.position() / self.scale
+            self.lastPoint = zoom_corrected_pos
+            self.drawOnImage(event)
+        else:
+            self.drawing = False
+
+
+    # can I pass this event to something else to synchronize the mouse moves
+    def mouseMoveEvent(self, event):
+        if not self.hasMouseTracking() or not self.drawing_enabled:
+            return
+        else:
+            if self.propagate_mouse_move == True:
+                self.propagate_mouse_event(event)
+
+        # print('in mouse move', self.hasMouseTracking(), self.drawing, self.vdp.active)
+        if self.statusBar:
+            zoom_corrected_pos = event.position() / self.scale
+            self.statusBar.showMessage('x=' + str(zoom_corrected_pos.x()) + ' y=' + str(
+                zoom_corrected_pos.y()))
+        if self.vdp.active:
+            self.vdp.mouseMoveEvent(event)
+            region = self.scrollArea.widget().visibleRegion()
+            self.update(region)
+            return
+        self.drawOnImage(event)
+
+    def mouseReleaseEvent(self, event):
+        self.drawing = False
+        if not self.hasMouseTracking() or not self.drawing_enabled:
+            return
+        if self.vdp.active:
+            self.vdp.mouseReleaseEvent(event)
+            self.update()  # required to update drawing
+            return
+        # if event.button == QtCore.Qt.LeftButton:
+        #     self.drawing = False
+        if self.clickCount == 1:
+            QTimer.singleShot(QApplication.instance().doubleClickInterval(),
+                              self.updateButtonCount)
+
+    def propagate_mouse_event(self, event):
+        print(self,event)
+        pass
+
+    def drawOnImage(self, event):
+        # bug fix for image drawing when image is none
+        if self.imageDraw is None:
+            return
+        zoom_corrected_pos = event.position() / self.scale
+        if self.drawing and (event.buttons() == QtCore.Qt.LeftButton or event.buttons() == QtCore.Qt.RightButton):
+            # now drawing or erasing over the image
+            self._draw_on_image(self.imageDraw, event, zoom_corrected_pos)
+            # this contains only the user drawing and nothing else --> can be useful to have too!!!
+            self._draw_on_image(self.raw_user_drawing, event, zoom_corrected_pos, erase_color=self.eraseColor_visualizer)
+
+
+
+        # Drawing the cursor TODO add boolean to ask if drawing cursor should be shown
+        painter = QtGui.QPainter(self.cursor)
+        try:
+            # We erase previous pointer
+            r = QtCore.QRect(QtCore.QPoint(), self._clear_size * QtCore.QSize() * self.brushSize) # WHY DO I START FROM 0, 0 (empty point I should have taken old coord of self.lastPoint!!! MEGA TODO
+            painter.save()
+            r.moveCenter(self.lastPoint.toPoint())
+            painter.setCompositionMode(QtGui.QPainter.CompositionMode_Clear)
+            painter.eraseRect(r)
+            painter.restore()
+            # draw the new one
+
+            stroke_size = 2
+            if self.brushSize<6:
+                stroke_size = 1
+
+            # if is 1 --> I should really draw a point instead
+            # I should also hide the cursor --> that would help
+            painter.setPen(QtGui.QPen(self.cursorColor, stroke_size, QtCore.Qt.SolidLine, QtCore.Qt.RoundCap,
+                                      QtCore.Qt.RoundJoin))
+            if self.brushSize >1:
+                painter.drawEllipse(zoom_corrected_pos, int(self.brushSize / 2.),
+                                int(self.brushSize / 2.))
+            else:
+                painter.drawPoint(zoom_corrected_pos)
+        except:
+            traceback.print_exc()
+        finally:
+            painter.end()
+        try:
+            region = self.scrollArea.widget().visibleRegion()
+        except:
+            # assume no scroll area --> use self visible region
+            region =self.visibleRegion()
+
+        self.update(region)
+
+        # required to erase mouse pointer
+        self.lastPoint = zoom_corrected_pos
+
+
+    def _draw_on_image(self, image_to_be_drawn, event, zoom_corrected_pos, erase_color=None):
+        painter = QtGui.QPainter(image_to_be_drawn)
+        try:
+            # small hack to allow use pen or fingers on touch screens to erase when Ctrl or shift is pressed along with the click --> MAYBE PUT THIS AS AN OPTION SOME DAY BECAUSE THIS MAY PERTURB PEOPLE
+            if event.buttons() == QtCore.Qt.LeftButton and not (event.modifiers() == getCtrlModifier() or event.modifiers() ==QtCore.Qt.ShiftModifier):
+                painter.setPen(QtGui.QPen(self.drawColor, self.brushSize, QtCore.Qt.SolidLine, QtCore.Qt.RoundCap,
+                                          QtCore.Qt.RoundJoin))
+            else:
+                if erase_color is None:
+                    erase_color = self.eraseColor
+                painter.setPen(QtGui.QPen(erase_color, self.brushSize, QtCore.Qt.SolidLine, QtCore.Qt.RoundCap,
+                                          QtCore.Qt.RoundJoin))
+            if self.lastPoint != zoom_corrected_pos:
+                painter.drawLine(self.lastPoint, zoom_corrected_pos)
+            else:
+                # if zero length line then draw point instead
+                painter.drawPoint(zoom_corrected_pos)
+        except:
+            traceback.print_exc()
+        finally:
+            painter.end()
+
+    # adds context/right click menu but only in vectorial mode
+    def contextMenuEvent(self, event):
+        if not self.vdp.active:
+            return
+        cmenu = QMenu(self)
+        newAct = cmenu.addAction("New")
+        opnAct = cmenu.addAction("Open")
+        quitAct = cmenu.addAction("Quit")
+        action = cmenu.exec_(self.mapToGlobal(event.pos()))
+        if action == quitAct:
+            sys.exit(0)
+
+    def updateButtonCount(self):
+        self.clickCount = 1
+
+    def mouseDoubleClickEvent(self, event):
+        self.clickCount = 2
+
+        # TODO maybe add controls to be sure we are in drawing mode with a pen ???
+        # double click fill or delete
+        if not self.vdp.active:
+            zoom_corrected_pos = event.position() / self.scale
+            if event.buttons() == QtCore.Qt.LeftButton:
+                # double click fill
+                if self.dialog_fill_erase('fill'):
+                    self.double_click_fill(zoom_corrected_pos)
+            elif event.buttons() == QtCore.Qt.RightButton:
+                # double click delete shape
+                if self.dialog_fill_erase('erase'):
+                    self.double_click_erase(zoom_corrected_pos)
+
+        self.vdp.mouseDoubleClickEvent(event)
+
+    def dialog_fill_erase(self, action='fill'):
+        choice = QMessageBox.question(self, 'Double click detected!',
+                                      'Do you want to '+ action +' the shape?',
+                                      QMessageBox.Yes | QMessageBox.No)
+        if choice == QMessageBox.Yes:
+            return True
+        else:
+            return False
+
+
+    def double_click_fill(self, zoom_corrected_pos):
+        try:
+            # erased, drawn = self.get_user_drawing(show_erased=True)
+            usr_drawing = self.get_user_drawing()
+            labs = label(usr_drawing, connectivity=2, background=0)
+            # do I really need that can I rather do np.unique or alike and will it be faster
+            rps_user_drawing = regionprops(labs)
+            if len(rps_user_drawing) == 0:
+                return
+            drawn_mask = self.get_mask()
+            # cell clicked over
+            cell_id_clicked_over = labs[int(zoom_corrected_pos.y()), int(zoom_corrected_pos.x())]
+            if cell_id_clicked_over == 0:
+                return
+            # we remove the cell drawn over
+            drawn_mask[labs==cell_id_clicked_over] = 0
+            # maintenant j'inverse et je trouve la cellule à filler sur le negatif
+            lab_cells_user_drawing = label(drawn_mask, connectivity=1, background=255)
+            cell_id_clicked_over2 = lab_cells_user_drawing[int(zoom_corrected_pos.y()), int(zoom_corrected_pos.x())]
+            drawn_mask[lab_cells_user_drawing==cell_id_clicked_over2]=255
+            self.set_mask(drawn_mask)
+        except:
+            logger.error('Double click fill failed')
+            traceback.print_exc()
+
+
+    def double_click_erase(self, zoom_corrected_pos):
+        try:
+            # erased, drawn = self.get_user_drawing(show_erased=True)
+            erased, drawn = self.get_user_drawing(show_erased=True)
+            labs = label(erased, connectivity=2, background=0)
+
+            # do I really need that can I rather do np.unique or alike and will it be faster
+            rps_user_drawing = regionprops(labs)
+            if len(rps_user_drawing) == 0:
+                return
+            drawn_mask = self.get_mask()
+            # cell clicked over
+            cell_id_clicked_over = labs[int(zoom_corrected_pos.y()), int(zoom_corrected_pos.x())]
+            if cell_id_clicked_over==0:
+                return
+            # we refill the removed cell
+            drawn_mask[labs==cell_id_clicked_over] = 255
+            # maintenant j'inverse et je trouve la cellule à filler sur le negatif
+            lab_cells_user_drawing = label(drawn_mask, connectivity=1, background=0)
+            cell_id_clicked_over2 = lab_cells_user_drawing[int(zoom_corrected_pos.y()), int(zoom_corrected_pos.x())]
+            drawn_mask[lab_cells_user_drawing==cell_id_clicked_over2]=0
+
+
+            self.set_mask(drawn_mask)
+        except:
+            logger.error('Double click delete failed')
+            traceback.print_exc()
+
+
+    def paintEvent(self, event):
+        # super(Createpaintwidget, self).paintEvent(event) # KEEP somehow tjis causes bugs --> do not uncomment it
+        canvasPainter = QtGui.QPainter(self)
+        try:
+
+            # the scrollpane visible region
+            try:
+                visibleRegion = self.scrollArea.widget().visibleRegion()
+            except:
+                #assume no scroll region --> visible region is self visible region
+                visibleRegion = self.visibleRegion()
+            # the corresponding rect
+            visibleRect = visibleRegion.boundingRect()
+            # the visibleRect taking zoom into account
+            scaledVisibleRect = QRect(int(visibleRect.x() / self.scale), int(visibleRect.y() / self.scale),
+                                      int(visibleRect.width() / self.scale), int(visibleRect.height() / self.scale))
+
+            if self.image is None:
+                canvasPainter.eraseRect(visibleRect)
+                # canvasPainter.end()
+                return
+
+            canvasPainter.drawImage(visibleRect, self.image, scaledVisibleRect)
+            if not self.vdp.active and self.maskVisible and self.imageDraw is not None:
+                canvasPainter.drawImage(visibleRect, self.imageDraw, scaledVisibleRect)
+                # should draw the cursor
+            canvasPainter.drawImage(visibleRect, self.cursor, scaledVisibleRect)
+
+            # ça ajoute un texte --> pas mal mais le refresh on scroll est pas top --> dois je le dessiner plutot dans vdp --> another set of objects that are non interactible and also cannot be saved and local???
+
+            # if True:
+            #     print('adding text')
+            #     myQPen = QPen(QtCore.Qt.blue, 3)
+            #     canvasPainter.setPen(myQPen)
+            #     canvasPainter.setPen(QColor("Green"))
+            #     canvasPainter.setFont(QFont('Helvetica', 48))
+            #     #     print('drawing text')
+            #     canvasPainter.drawText(visibleRect.x(), visibleRect.y() +10, 'this is a test of your system')
+            #     # pb here is that the images is not cleared --> creates bugs
+
+            if self.vdp.active:
+                self.vdp.paintEvent(canvasPainter, scaledVisibleRect)
+        except:
+            traceback.print_exc()
+        finally:
+            canvasPainter.end()
+
+if __name__ == '__main__':
+    # TODO add a main method so it can be called directly
+    # maybe just show a canvas and give it interesting props --> TODO --> really need fix that too!!!
+    import sys
+    from qtpy.QtWidgets import QApplication
+    from matplotlib import pyplot as plt
+
+    # should probably have his own scroll bar embedded somewhere
+
+    app = QApplication(sys.argv)
+
+
+    # in fact that is easy to override methods before construction --> can easily do that in fact
+
+    # quite easy todo in fact
+    class overriding_apply(Createpaintwidget):
+        def apply(self):
+            plt.imshow(self.get_mask())
+            plt.show()
+
+    #Do not use show() in a GUI application.
+    # If you have a FigureCanvas instance embedded in your app, call its
+    # draw() method.
+    # If you use pyplot.figure() to create a matplotlib window from your
+    # app, call pyplot.draw().
+
+    # maybe add a figure canvas the to my code and plot to it and draw it --> quite easy and nice plotting capabilities then in python
+
+
+
+    # w = Createpaintwidget(enable_shortcuts=True)
+    w = overriding_apply(enable_shortcuts=True)
+    # w.set_image('/E/Sample_images/sample_images_PA/mini (copie)/focused_Series012.png')
+    w.set_image('/E/Sample_images/sample_images_PA/mini/focused_Series012.png')
+    # w.set_image('/E/Sample_images/wings/fluorescent_wings_spots_charroux/909dsRed/0.tif')
+    # w.set_image('/E/Sample_images/wings/fluorescent_wings_spots_charroux/contoles_test_X1VK06/0.tif')
+    # w.set_mask('/E/Sample_images/sample_images_PA/mini (copie)/focused_Series012/handCorrection.png')
+    # w.set_image(Img('/E/Sample_images/sample_images_PA/mini (copie)/focused_Series012.png'))
+
+    # print(tst.max())
+    # tst = tst[..., np.newaxis]
+
+
+
+
+    # tst = Img('/E/Sample_images/wings/fluorescent_wings_spots_charroux/909dsRed/0.tif')
+    # tst = Img('/E/Sample_images/wings/fluorescent_wings_spots_charroux/contoles_test_X1VK06/0.tif')
+    # tst = Img('/E/Sample_images/sample_images_PA/mini (copie)/focused_Series012.png')
+    # tst = normalization(tst, method='Rescaling (min-max normalization)', range=[0,1], individual_channels=len(tst.shape)>2, clip=True)
+    # tst = normalization(tst, method='Rescaling (min-max normalization)', range=[0,1], individual_channels=len(tst.shape)>2, clip=True)
+    # tst = normalization(tst, method='Standardization', range=[0,1], individual_channels=True, clip=True)
+    # tst = normalization(tst, method='Percentile', range=[0,1], individual_channels=len(tst.shape)>2, clip=True, normalization_minima_and_maxima=[1.,50.0])
+    # tst = np.squeeze(tst)
+
+    # tst = auto_threshold(tst)
+    # save_as_tiff(tst, output_name='/E/Sample_images/sample_images_PA/mini (copie)/tst_norm.tif')
+
+    # do I have a standar
+    # tst = normalization(tst, method='standardization', range=[0,1], individual_channels=True)
+    # print(tst.max())
+    # w.set_image(tst)
+
+    # all is so easy to do this way
+
+
+    # ça marche --> tres facile a tester
+    # --> how can I do a zoom
+    # ask whether self scroll or not --> if scrollable
+
+    # mask = w.get_mask()
+    #
+    # plt.imshow(mask)
+    # plt.show()
+
+    w.show()
+    sys.exit(app.exec_())
+