"""
ImageView.py -  Widget for basic image dispay and analysis
Copyright 2010  Luke Campagnola
Distributed under MIT/X11 license. See license.txt for more information.

Widget used for displaying 2D or 3D data. Features:
  - float or int (including 16-bit int) image display via ImageItem
  - zoom/pan via GraphicsView
  - black/white level controls
  - time slider for 3D data sets
  - ROI plotting
  - Image normalization through a variety of methods
"""
import os
from math import log10
from time import perf_counter

import numpy as np

from .. import debug as debug
from .. import functions as fn
from .. import getConfigOption
from ..graphicsItems.GradientEditorItem import addGradientListToDocstring
from ..graphicsItems.ImageItem import ImageItem
from ..graphicsItems.InfiniteLine import InfiniteLine
from ..graphicsItems.LinearRegionItem import LinearRegionItem
from ..graphicsItems.ROI import ROI
from ..graphicsItems.ViewBox import ViewBox
from ..graphicsItems.VTickGroup import VTickGroup
from ..Qt import QtCore, QtGui, QtWidgets
from ..SignalProxy import SignalProxy
from . import ImageViewTemplate_generic as ui_template

try:
    from bottleneck import nanmax, nanmin
except ImportError:
    from numpy import nanmax, nanmin

translate = QtCore.QCoreApplication.translate


class PlotROI(ROI):
    def __init__(self, size):
        ROI.__init__(self, pos=[0,0], size=size) #, scaleSnap=True, translateSnap=True)
        self.addScaleHandle([1, 1], [0, 0])
        self.addRotateHandle([0, 0], [0.5, 0.5])


class ImageView(QtWidgets.QWidget):
    """
    Widget used for display and analysis of image data.
    Implements many features:
    
      * Displays 2D and 3D image data. For 3D data, a z-axis
        slider is displayed allowing the user to select which frame is displayed.
      * Displays histogram of image data with movable region defining the dark/light levels
      * Editable gradient provides a color lookup table
      * Frame slider may also be moved using left/right arrow keys as well as pgup, pgdn, home, and end.
      * Basic analysis features including:

          * ROI and embedded plot for measuring image values across frames
          * Image normalization / background subtraction
    
    Basic Usage::
    
        imv = pg.ImageView()
        imv.show()
        imv.setImage(data)
        
    **Keyboard interaction**
    
      * left/right arrows step forward/backward 1 frame when pressed,
        seek at 20fps when held.
      * up/down arrows seek at 100fps
      * pgup/pgdn seek at 1000fps
      * home/end seek immediately to the first/last frame
      * space begins playing frames. If time values (in seconds) are given
        for each frame, then playback is in realtime.
    """
    sigTimeChanged = QtCore.Signal(object, object)
    sigProcessingChanged = QtCore.Signal(object)
    
    def __init__(
            self,
            parent=None,
            name="ImageView",
            view=None,
            imageItem=None,
            levelMode='mono',
            discreteTimeLine=False,
            roi=None,
            normRoi=None,
            *args,
    ):
        """
        By default, this class creates an :class:`ImageItem <pyqtgraph.ImageItem>` to display image data
        and a :class:`ViewBox <pyqtgraph.ViewBox>` to contain the ImageItem.

        Parameters
        ----------
        parent : QWidget
            Specifies the parent widget to which this ImageView will belong. If None, then the ImageView is created with
            no parent.
        name : str
            The name used to register both the internal ViewBox and the PlotItem used to display ROI data. See the
            *name* argument to :func:`ViewBox.__init__() <pyqtgraph.ViewBox.__init__>`.
        view : ViewBox or PlotItem
            If specified, this will be used as the display area that contains the displayed image. Any
            :class:`ViewBox <pyqtgraph.ViewBox>`, :class:`PlotItem <pyqtgraph.PlotItem>`, or other compatible object is
            acceptable. Note: to display axis ticks inside the ImageView, instantiate it with a PlotItem instance as its
            view::

                pg.ImageView(view=pg.PlotItem())
        imageItem : ImageItem
            If specified, this object will be used to display the image. Must be an instance of ImageItem or other
            compatible object.
        levelMode : str
            See the *levelMode* argument to :func:`HistogramLUTItem.__init__() <pyqtgraph.HistogramLUTItem.__init__>`
        discreteTimeLine : bool
            Whether to snap to xvals / frame numbers when interacting with the timeline position.
        roi : ROI
            If specified, this object is used as ROI for the plot feature. Must be an instance of ROI.
        normRoi : ROI
            If specified, this object is used as ROI for the normalization feature. Must be an instance of ROI.
        """
        QtWidgets.QWidget.__init__(self, parent, *args)
        self._imageLevels = None  # [(min, max), ...] per channel image metrics
        self.levelMin = None    # min / max levels across all channels
        self.levelMax = None

        self.name = name
        self.image = None
        self.axes = {}
        self.imageDisp = None
        self.ui = ui_template.Ui_Form()
        self.ui.setupUi(self)
        self.scene = self.ui.graphicsView.scene()
        self.discreteTimeLine = discreteTimeLine
        self.ui.histogram.setLevelMode(levelMode)
        self.ignoreTimeLine = False

        if view is None:
            self.view = ViewBox()
        else:
            self.view = view
        self.ui.graphicsView.setCentralItem(self.view)
        self.view.setAspectLocked(True)
        self.view.invertY()
        
        self.menu = None
        
        self.ui.normGroup.hide()

        if roi is None:
            self.roi = PlotROI(10)
        else:
            self.roi = roi
        self.roi.setZValue(20)
        self.view.addItem(self.roi)
        self.roi.hide()
        if normRoi is None:
            self.normRoi = PlotROI(10)
            self.normRoi.setPen('y')
        else:
            self.normRoi = normRoi
        self.normRoi.setZValue(20)
        self.view.addItem(self.normRoi)
        self.normRoi.hide()
        self.roiCurves = []
        self.timeLine = InfiniteLine(0, movable=True)
        if getConfigOption('background')=='w':
            self.timeLine.setPen((20, 80,80, 200))
        else:
            self.timeLine.setPen((255, 255, 0, 200))
        self.timeLine.setZValue(1)
        self.ui.roiPlot.addItem(self.timeLine)
        self.ui.splitter.setSizes([self.height()-35, 35])

        # init imageItem and histogram
        if imageItem is None:
            self.imageItem = ImageItem()
        else:
            self.imageItem = imageItem
            self.setImage(imageItem.image, autoRange=False, autoLevels=False, transform=imageItem.transform())
        self.view.addItem(self.imageItem)
        self.currentIndex = 0
        
        self.ui.histogram.setImageItem(self.imageItem)
        self.ui.histogram.setLevelMode(levelMode)
        
        # make splitter an unchangeable small grey line:
        s = self.ui.splitter
        s.handle(1).setEnabled(False)
        s.setStyleSheet("QSplitter::handle{background-color: grey}")
        s.setHandleWidth(2)

        self.ui.roiPlot.hideAxis('left')
        self.frameTicks = VTickGroup(yrange=[0.8, 1], pen=0.4)
        self.ui.roiPlot.addItem(self.frameTicks, ignoreBounds=True)
        
        self.keysPressed = {}
        self.playTimer = QtCore.QTimer()
        self.playRate = 0
        self._pausedPlayRate = None
        self.fps = 1  # 1 Hz by default
        self.lastPlayTime = 0
        
        self.normRgn = LinearRegionItem()
        self.normRgn.setZValue(0)
        self.ui.roiPlot.addItem(self.normRgn)
        self.normRgn.hide()
            
        ## wrap functions from view box
        for fn in ['addItem', 'removeItem']:
            setattr(self, fn, getattr(self.view, fn))

        ## wrap functions from histogram
        for fn in ['setHistogramRange', 'autoHistogramRange', 'getLookupTable', 'getLevels']:
            setattr(self, fn, getattr(self.ui.histogram, fn))

        self.timeLine.sigPositionChanged.connect(self.timeLineChanged)
        self.ui.roiBtn.clicked.connect(self.roiClicked)
        self.roi.sigRegionChanged.connect(self.roiChanged)
        #self.ui.normBtn.toggled.connect(self.normToggled)
        self.ui.menuBtn.clicked.connect(self.menuClicked)
        self.ui.normDivideRadio.clicked.connect(self.normRadioChanged)
        self.ui.normSubtractRadio.clicked.connect(self.normRadioChanged)
        self.ui.normOffRadio.clicked.connect(self.normRadioChanged)
        self.ui.normROICheck.clicked.connect(self.updateNorm)
        self.ui.normFrameCheck.clicked.connect(self.updateNorm)
        self.ui.normTimeRangeCheck.clicked.connect(self.updateNorm)
        self.playTimer.timeout.connect(self.timeout)
        
        self.normProxy = SignalProxy(
            self.normRgn.sigRegionChanged,
            slot=self.updateNorm,
            threadSafe=False,
        )
        self.normRoi.sigRegionChangeFinished.connect(self.updateNorm)
        
        self.ui.roiPlot.registerPlot(self.name + '_ROI')
        self.view.register(self.name)
        
        self.noRepeatKeys = [
            QtCore.Qt.Key.Key_Right,
            QtCore.Qt.Key.Key_Left,
            QtCore.Qt.Key.Key_Up,
            QtCore.Qt.Key.Key_Down,
            QtCore.Qt.Key.Key_PageUp,
            QtCore.Qt.Key.Key_PageDown,
        ]
        
        self.roiClicked() ## initialize roi plot to correct shape / visibility

    def setImage(
            self,
            img,
            autoRange=True,
            autoLevels=True,
            levels=None,
            axes=None,
            xvals=None,
            pos=None,
            scale=None,
            transform=None,
            autoHistogramRange=True,
            levelMode=None,
    ):
        """
        Set the image to be displayed in the widget.

        Parameters
        ----------
        img : np.ndarray
            The image to be displayed. See :func:`ImageItem.setImage` and *notes* below.
        autoRange : bool
            Whether to scale/pan the view to fit the image.
        autoLevels : bool
            Whether to update the white/black levels to fit the image.
        levels : tuple
            (min, max) white and black level values to use.
        axes : dict
            Dictionary indicating the interpretation for each axis. This is only needed to override the default guess.
            Format is::

                {'t':0, 'x':1, 'y':2, 'c':3};
        xvals : np.ndarray
            1D array of values corresponding to the first axis in a 3D image. For video, this array should contain
            the time of each frame.
        pos
            Change the position of the displayed image
        scale
            Change the scale of the displayed image
        transform
            Set the transform of the displayed image. This option overrides *pos* and *scale*.
        autoHistogramRange : bool
            If True, the histogram y-range is automatically scaled to fit the image data.
        levelMode : str
            If specified, this sets the user interaction mode for setting image levels. Options are 'mono',
            which provides a single level control for all image channels, and 'rgb' or 'rgba', which provide
            individual controls for each channel.

        Notes
        -----
        For backward compatibility, image data is assumed to be in column-major order (column, row).
        However, most image data is stored in row-major order (row, column) and will need to be
        transposed before calling setImage()::
        
            imageview.setImage(imagedata.T)
            
        This requirement can be changed by the ``imageAxisOrder``
        :ref:`global configuration option <apiref_config>`.
        """
        profiler = debug.Profiler()

        if hasattr(img, 'implements') and img.implements('MetaArray'):
            img = img.asarray()

        if not isinstance(img, np.ndarray):
            required = ['dtype', 'max', 'min', 'ndim', 'shape', 'size']
            if not all(hasattr(img, attr) for attr in required):
                raise TypeError("Image must be NumPy array or any object "
                                "that provides compatible attributes/methods:\n"
                                "  %s" % str(required))

        self.image = img
        self.imageDisp = None
        if levelMode is not None:
            self.ui.histogram.setLevelMode(levelMode)

        profiler()

        if axes is None:
            x,y = (0, 1) if self.imageItem.axisOrder == 'col-major' else (1, 0)

            if img.ndim == 2:
                self.axes = {'t': None, 'x': x, 'y': y, 'c': None}
            elif img.ndim == 3:
                # Ambiguous case; make a guess
                if img.shape[2] <= 4:
                    self.axes = {'t': None, 'x': x, 'y': y, 'c': 2}
                else:
                    self.axes = {'t': 0, 'x': x+1, 'y': y+1, 'c': None}
            elif img.ndim == 4:
                # Even more ambiguous; just assume the default
                self.axes = {'t': 0, 'x': x+1, 'y': y+1, 'c': 3}
            else:
                raise Exception("Can not interpret image with dimensions %s" % (str(img.shape)))
        elif isinstance(axes, dict):
            self.axes = axes.copy()
        elif isinstance(axes, list) or isinstance(axes, tuple):
            self.axes = {}
            for i in range(len(axes)):
                self.axes[axes[i]] = i
        else:
            raise Exception("Can not interpret axis specification %s. Must be like {'t': 2, 'x': 0, 'y': 1} or ('t', 'x', 'y', 'c')" % (str(axes)))
            
        for x in ['t', 'x', 'y', 'c']:
            self.axes[x] = self.axes.get(x, None)
        axes = self.axes

        if xvals is not None:
            self.tVals = xvals
        elif axes['t'] is not None:
            if hasattr(img, 'xvals'):
                try:
                    self.tVals = img.xvals(axes['t'])
                except:
                    self.tVals = np.arange(img.shape[axes['t']])
            else:
                self.tVals = np.arange(img.shape[axes['t']])

        profiler()

        self.currentIndex = 0
        self.updateImage(autoHistogramRange=autoHistogramRange)
        if levels is None and autoLevels:
            self.autoLevels()
        if levels is not None:  ## this does nothing since getProcessedImage sets these values again.
            self.setLevels(*levels)
            
        if self.ui.roiBtn.isChecked():
            self.roiChanged()

        profiler()

        if self.axes['t'] is not None:
            self.ui.roiPlot.setXRange(self.tVals.min(), self.tVals.max())
            self.frameTicks.setXVals(self.tVals)
            self.timeLine.setValue(0)
            if len(self.tVals) > 1:
                start = self.tVals.min()
                stop = self.tVals.max() + abs(self.tVals[-1] - self.tVals[0]) * 0.02
            elif len(self.tVals) == 1:
                start = self.tVals[0] - 0.5
                stop = self.tVals[0] + 0.5
            else:
                start = 0
                stop = 1
            for s in [self.timeLine, self.normRgn]:
                s.setBounds([start, stop])
        
        profiler()

        if transform is None:
            transform = QtGui.QTransform()
            # note that the order of transform is
            #   scale followed by translate
            if pos is not None:
                transform.translate(*pos)
            if scale is not None:
                transform.scale(*scale)
        self.imageItem.setTransform(transform)

        profiler()

        if autoRange:
            self.autoRange()
        self.roiClicked()

        profiler()

    def clear(self):
        self.image = None
        self.imageItem.clear()
        
    def play(self, rate=None):
        """Begin automatically stepping frames forward at the given rate (in fps).
        This can also be accessed by pressing the spacebar."""
        if rate is None:
            rate = self._pausedPlayRate or self.fps
        if rate == 0 and self.playRate not in (None, 0):
            self._pausedPlayRate = self.playRate
        self.playRate = rate

        if rate == 0:
            self.playTimer.stop()
            return
            
        self.lastPlayTime = perf_counter()
        if not self.playTimer.isActive():
            self.playTimer.start(abs(int(1000/rate)))

    def togglePause(self):
        if self.playTimer.isActive():
            self.play(0)
        elif self.playRate == 0:
            if self._pausedPlayRate is not None:
                fps = self._pausedPlayRate
            else:
                fps = (self.nframes() - 1) / (self.tVals[-1] - self.tVals[0])
            self.play(fps)
        else:
            self.play(self.playRate)

    def setHistogramLabel(self, text=None, **kwargs):
        """
        Set the label text of the histogram axis similar to
        :func:`AxisItem.setLabel() <pyqtgraph.AxisItem.setLabel>`
        """
        a = self.ui.histogram.axis
        a.setLabel(text, **kwargs)
        if text == '':
            a.showLabel(False)
        self.ui.histogram.setMinimumWidth(135)

    def nframes(self):
        """
        Returns
        -------
        int
            The number of frames in the image data.
        """
        if self.image is None:
            return 0
        elif self.axes['t'] is not None:
            return self.image.shape[self.axes['t']]
        return 1

    def autoLevels(self):
        """Set the min/max intensity levels automatically to match the image data."""
        self.setLevels(rgba=self._imageLevels)

    def setLevels(self, *args, **kwds):
        """Set the min/max (bright and dark) levels.
        
        See :func:`HistogramLUTItem.setLevels <pyqtgraph.HistogramLUTItem.setLevels>`.
        """
        self.ui.histogram.setLevels(*args, **kwds)

    def autoRange(self):
        """Auto scale and pan the view around the image such that the image fills the view."""
        self.getProcessedImage()
        self.view.autoRange()
        
    def getProcessedImage(self):
        """Returns the image data after it has been processed by any normalization options in use.
        """
        if self.imageDisp is None:
            image = self.normalize(self.image)
            self.imageDisp = image
            self._imageLevels = self.quickMinMax(self.imageDisp)
            self.levelMin = min([level[0] for level in self._imageLevels])
            self.levelMax = max([level[1] for level in self._imageLevels])
            
        return self.imageDisp
        
    def close(self):
        """Closes the widget nicely, making sure to clear the graphics scene and release memory."""
        self.clear()
        self.imageDisp = None
        self.imageItem.setParent(None)
        super(ImageView, self).close()
        self.setParent(None)
        
    def keyPressEvent(self, ev):
        if not self.hasTimeAxis():
            super().keyPressEvent(ev)
            return

        if ev.key() == QtCore.Qt.Key.Key_Space:
            self.togglePause()
            ev.accept()
        elif ev.key() == QtCore.Qt.Key.Key_Home:
            self.setCurrentIndex(0)
            self.play(0)
            ev.accept()
        elif ev.key() == QtCore.Qt.Key.Key_End:
            self.setCurrentIndex(self.nframes()-1)
            self.play(0)
            ev.accept()
        elif ev.key() in self.noRepeatKeys:
            ev.accept()
            if ev.isAutoRepeat():
                return
            self.keysPressed[ev.key()] = 1
            self.evalKeyState()
        else:
            super().keyPressEvent(ev)

    def keyReleaseEvent(self, ev):
        if not self.hasTimeAxis():
            super().keyReleaseEvent(ev)
            return

        if ev.key() in [QtCore.Qt.Key.Key_Space, QtCore.Qt.Key.Key_Home, QtCore.Qt.Key.Key_End]:
            ev.accept()
        elif ev.key() in self.noRepeatKeys:
            ev.accept()
            if ev.isAutoRepeat():
                return
            try:
                del self.keysPressed[ev.key()]
            except:
                self.keysPressed = {}
            self.evalKeyState()
        else:
            super().keyReleaseEvent(ev)
        
    def evalKeyState(self):
        if len(self.keysPressed) == 1:
            key = list(self.keysPressed.keys())[0]
            if key == QtCore.Qt.Key.Key_Right:
                self.play(20)
                self.jumpFrames(1)
                # effectively pause playback for 0.2 s
                self.lastPlayTime = perf_counter() + 0.2  
            elif key == QtCore.Qt.Key.Key_Left:
                self.play(-20)
                self.jumpFrames(-1)
                self.lastPlayTime = perf_counter() + 0.2
            elif key == QtCore.Qt.Key.Key_Up:
                self.play(-100)
            elif key == QtCore.Qt.Key.Key_Down:
                self.play(100)
            elif key == QtCore.Qt.Key.Key_PageUp:
                self.play(-1000)
            elif key == QtCore.Qt.Key.Key_PageDown:
                self.play(1000)
        else:
            self.play(0)
        
    def timeout(self):
        now = perf_counter()
        dt = now - self.lastPlayTime
        if dt < 0:
            return
        n = int(self.playRate * dt)
        if n != 0:
            self.lastPlayTime += (float(n)/self.playRate)
            if self.currentIndex+n > self.image.shape[self.axes['t']]:
                self.play(0)
            self.jumpFrames(n)
        
    def setCurrentIndex(self, ind):
        """Set the currently displayed frame index."""
        index = fn.clip_scalar(ind, 0, self.nframes()-1)
        self.currentIndex = index
        self.updateImage()
        self.ignoreTimeLine = True
        # Implicitly call timeLineChanged
        self.timeLine.setValue(self.tVals[index])
        self.ignoreTimeLine = False

    def jumpFrames(self, n):
        """Move video frame ahead n frames (may be negative)"""
        if self.axes['t'] is not None:
            self.setCurrentIndex(self.currentIndex + n)

    def normRadioChanged(self):
        self.imageDisp = None
        self.updateImage()
        self.autoLevels()
        self.roiChanged()
        self.sigProcessingChanged.emit(self)
    
    def updateNorm(self):
        if self.ui.normTimeRangeCheck.isChecked():
            self.normRgn.show()
        else:
            self.normRgn.hide()
        
        if self.ui.normROICheck.isChecked():
            self.normRoi.show()
        else:
            self.normRoi.hide()
        
        if not self.ui.normOffRadio.isChecked():
            self.imageDisp = None
            self.updateImage()
            self.autoLevels()
            self.roiChanged()
            self.sigProcessingChanged.emit(self)

    def normToggled(self, b):
        self.ui.normGroup.setVisible(b)
        self.normRoi.setVisible(b and self.ui.normROICheck.isChecked())
        self.normRgn.setVisible(b and self.ui.normTimeRangeCheck.isChecked())

    def hasTimeAxis(self):
        return 't' in self.axes and self.axes['t'] is not None

    def roiClicked(self):
        showRoiPlot = False
        if self.ui.roiBtn.isChecked():
            showRoiPlot = True
            self.roi.show()
            self.ui.roiPlot.setMouseEnabled(True, True)
            self.ui.splitter.setSizes([int(self.height()*0.6), int(self.height()*0.4)])
            self.ui.splitter.handle(1).setEnabled(True)
            self.roiChanged()
            for c in self.roiCurves:
                c.show()
            self.ui.roiPlot.showAxis('left')
        else:
            self.roi.hide()
            self.ui.roiPlot.setMouseEnabled(False, False)
            for c in self.roiCurves:
                c.hide()
            self.ui.roiPlot.hideAxis('left')
            
        if self.hasTimeAxis():
            showRoiPlot = True
            mn = self.tVals.min()
            mx = self.tVals.max()
            self.ui.roiPlot.setXRange(mn, mx, padding=0.01)
            self.timeLine.show()
            self.timeLine.setBounds([mn, mx])
            if not self.ui.roiBtn.isChecked():
                self.ui.splitter.setSizes([self.height()-35, 35])
                self.ui.splitter.handle(1).setEnabled(False)
        else:
            self.timeLine.hide()
            
        self.ui.roiPlot.setVisible(showRoiPlot)

    def roiChanged(self):
        # Extract image data from ROI
        if self.image is None:
            return

        image = self.getProcessedImage()

        # getArrayRegion axes should be (x, y) of data array for col-major,
        # (y, x) for row-major
        # can't just transpose input because ROI is axisOrder aware
        colmaj = self.imageItem.axisOrder == 'col-major'
        if colmaj:
            axes = (self.axes['x'], self.axes['y'])
        else:
            axes = (self.axes['y'], self.axes['x'])

        data, coords = self.roi.getArrayRegion(
            image.view(np.ndarray), img=self.imageItem, axes=axes,
            returnMappedCoords=True)

        if data is None:
            return

        # Convert extracted data into 1D plot data
        if self.axes['t'] is None:
            # Average across y-axis of ROI
            data = data.mean(axis=self.axes['y'])

            # get coordinates along x axis of ROI mapped to range (0, roiwidth)
            if colmaj:
                coords = coords[:, :, 0] - coords[:, 0:1, 0]
            else:
                coords = coords[:, 0, :] - coords[:, 0, 0:1]
            xvals = (coords**2).sum(axis=0) ** 0.5
        else:
            # Average data within entire ROI for each frame
            data = data.mean(axis=axes)
            xvals = self.tVals

        # Handle multi-channel data
        if data.ndim == 1:
            plots = [(xvals, data, 'w')]
        if data.ndim == 2:
            if data.shape[1] == 1:
                colors = 'w'
            else:
                colors = 'rgbw'
            plots = []
            for i in range(data.shape[1]):
                d = data[:,i]
                plots.append((xvals, d, colors[i]))

        # Update plot line(s)
        while len(plots) < len(self.roiCurves):
            c = self.roiCurves.pop()
            c.scene().removeItem(c)
        while len(plots) > len(self.roiCurves):
            self.roiCurves.append(self.ui.roiPlot.plot())
        for i in range(len(plots)):
            x, y, p = plots[i]
            self.roiCurves[i].setData(x, y, pen=p)

    def quickMinMax(self, data):
        """
        Estimate the min/max values of *data* by subsampling.
        Returns [(min, max), ...] with one item per channel
        """
        while data.size > 1e6:
            ax = np.argmax(data.shape)
            sl = [slice(None)] * data.ndim
            sl[ax] = slice(None, None, 2)
            data = data[tuple(sl)]
            
        cax = self.axes['c']
        if cax is None:
            if data.size == 0:
                return [(0, 0)]
            return [(float(nanmin(data)), float(nanmax(data)))]
        else:
            if data.size == 0:
                return [(0, 0)] * data.shape[-1]
            return [(float(nanmin(data.take(i, axis=cax))), 
                     float(nanmax(data.take(i, axis=cax)))) for i in range(data.shape[-1])]

    def normalize(self, image):
        """
        Process *image* using the normalization options configured in the
        control panel.
        
        This can be repurposed to process any data through the same filter.
        """
        if self.ui.normOffRadio.isChecked():
            return image
            
        div = self.ui.normDivideRadio.isChecked()
        norm = image.view(np.ndarray).copy()
        #if div:
            #norm = ones(image.shape)
        #else:
            #norm = zeros(image.shape)
        if div:
            norm = norm.astype(np.float32)
            
        if self.ui.normTimeRangeCheck.isChecked() and image.ndim == 3:
            (sind, start) = self.timeIndex(self.normRgn.lines[0])
            (eind, end) = self.timeIndex(self.normRgn.lines[1])
            #print start, end, sind, eind
            n = image[sind:eind+1].mean(axis=0)
            n.shape = (1,) + n.shape
            if div:
                norm /= n
            else:
                norm -= n
                
        if self.ui.normFrameCheck.isChecked() and image.ndim == 3:
            n = image.mean(axis=1).mean(axis=1)
            n.shape = n.shape + (1, 1)
            if div:
                norm /= n
            else:
                norm -= n
            
        if self.ui.normROICheck.isChecked() and image.ndim == 3:
            n = self.normRoi.getArrayRegion(norm, self.imageItem, (1, 2)).mean(axis=1).mean(axis=1)
            n = n[:,np.newaxis,np.newaxis]
            #print start, end, sind, eind
            if div:
                norm /= n
            else:
                norm -= n
                
        return norm
        
    def timeLineChanged(self):
        if not self.ignoreTimeLine:
            self.play(0)

        (ind, time) = self.timeIndex(self.timeLine)
        if ind != self.currentIndex:
            self.currentIndex = ind
            self.updateImage()
        if self.discreteTimeLine:
            with fn.SignalBlock(self.timeLine.sigPositionChanged, self.timeLineChanged):
                if self.tVals is not None:
                    self.timeLine.setPos(self.tVals[ind])
                else:
                    self.timeLine.setPos(ind)

        self.sigTimeChanged.emit(ind, time)

    def updateImage(self, autoHistogramRange=True):
        ## Redraw image on screen
        if self.image is None:
            return
    
        image = self.getProcessedImage()
        if autoHistogramRange:
            self.ui.histogram.setHistogramRange(self.levelMin, self.levelMax)
        
        # Transpose image into order expected by ImageItem
        if self.imageItem.axisOrder == 'col-major':
            axorder = ['t', 'x', 'y', 'c']
        else:
            axorder = ['t', 'y', 'x', 'c']
        axorder = [self.axes[ax] for ax in axorder if self.axes[ax] is not None]
        image = image.transpose(axorder)
            
        # Select time index
        if self.axes['t'] is not None:
            self.ui.roiPlot.show()
            image = image[self.currentIndex]
            
        self.imageItem.updateImage(image)

    def timeIndex(self, slider):
        """
        Returns
        -------
        int
            The index of the frame closest to the timeline slider.
        float
            The time value of the slider.
        """
        if not self.hasTimeAxis():
            return 0, 0.0

        t = slider.value()

        xv = self.tVals
        if xv is None:
            ind = int(t)
        else:
            if len(xv) < 2:
                return 0, 0.0
            inds = np.argwhere(xv <= t)
            if len(inds) < 1:
                return 0, t
            ind = inds[-1, 0]
        return ind, t

    def getView(self):
        """Return the ViewBox (or other compatible object) which displays the ImageItem"""
        return self.view
        
    def getImageItem(self):
        """Return the ImageItem for this ImageView."""
        return self.imageItem
        
    def getRoiPlot(self):
        """Return the ROI PlotWidget for this ImageView"""
        return self.ui.roiPlot
       
    def getHistogramWidget(self):
        """Return the HistogramLUTWidget for this ImageView"""
        return self.ui.histogram

    def export(self, fileName):
        """
        Export data from the ImageView to a file, or to a stack of files if
        the data is 3D. Saving an image stack will result in index numbers
        being added to the file name. Images are saved as they would appear
        onscreen, with levels and lookup table applied.
        """
        img = self.getProcessedImage()
        if self.hasTimeAxis():
            base, ext = os.path.splitext(fileName)
            fmt = "%%s%%0%dd%%s" % int(log10(img.shape[0])+1)
            for i in range(img.shape[0]):
                self.imageItem.setImage(img[i], autoLevels=False)
                self.imageItem.save(fmt % (base, i, ext))
            self.updateImage()
        else:
            self.imageItem.save(fileName)
            
    def exportClicked(self):
        fileName, _ = QtWidgets.QFileDialog.getSaveFileName()
        if not fileName:
            return
        self.export(fileName)
        
    def buildMenu(self):
        self.menu = QtWidgets.QMenu()
        self.normAction = QtGui.QAction(translate("ImageView", "Normalization"), self.menu)
        self.normAction.setCheckable(True)
        self.normAction.toggled.connect(self.normToggled)
        self.menu.addAction(self.normAction)
        self.exportAction = QtGui.QAction(translate("ImageView", "Export"), self.menu)
        self.exportAction.triggered.connect(self.exportClicked)
        self.menu.addAction(self.exportAction)
        
    def menuClicked(self):
        if self.menu is None:
            self.buildMenu()
        self.menu.popup(QtGui.QCursor.pos())

    def setColorMap(self, colormap):
        """Set the color map. 

        Parameters
        ----------
        colormap : ColorMap
            The ColorMap to use for coloring images.
        """
        self.ui.histogram.gradient.setColorMap(colormap)

    @addGradientListToDocstring()
    def setPredefinedGradient(self, name):
        """Set one of the gradients defined in :class:`GradientEditorItem`.
        Currently available gradients are:   
        """
        self.ui.histogram.gradient.loadPreset(name)