mitiff.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# Copyright (c) 2018, 2019 Satpy developers
#
# This file is part of satpy.
#
# satpy is free software: you can redistribute it and/or modify it under the
# terms of the GNU General Public License as published by the Free Software
# Foundation, either version 3 of the License, or (at your option) any later
# version.
#
# satpy is distributed in the hope that it will be useful, but WITHOUT ANY
# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
# A PARTICULAR PURPOSE.  See the GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License along with
# satpy.  If not, see <http://www.gnu.org/licenses/>.
"""MITIFF writer objects for creating MITIFF files from `Dataset` objects."""

import logging
import os

import dask
import numpy as np
from PIL import Image, ImagePalette

from satpy.dataset import DataID, DataQuery
from satpy.writers import ImageWriter, get_enhanced_image

IMAGEDESCRIPTION = 270

LOG = logging.getLogger(__name__)

KELVIN_TO_CELSIUS = -273.15


def _adjust_kwargs(dataset, kwargs):
    if 'platform_name' not in kwargs:
        kwargs['platform_name'] = dataset.attrs['platform_name']
    if 'name' not in kwargs:
        kwargs['name'] = dataset.attrs['name']
    if 'start_time' not in kwargs:
        kwargs['start_time'] = dataset.attrs['start_time']
    if 'sensor' not in kwargs:
        kwargs['sensor'] = dataset.attrs['sensor']
    # Sensor attrs could be set. MITIFFs needing to handle sensor can only have one sensor
    # Assume the first value of set as the sensor.
    if isinstance(kwargs['sensor'], set):
        LOG.warning('Sensor is set, will use the first value: %s', kwargs['sensor'])
        kwargs['sensor'] = (list(kwargs['sensor']))[0]


class MITIFFWriter(ImageWriter):
    """Writer to produce MITIFF image files."""

    def __init__(self, name=None, tags=None, **kwargs):
        """Initialize reader with tag and other configuration information."""
        ImageWriter.__init__(self, name=name, default_config_filename="writers/mitiff.yaml", **kwargs)
        self.tags = self.info.get("tags", None) if tags is None else tags
        if self.tags is None:
            self.tags = {}
        elif not isinstance(self.tags, dict):
            # if it's coming from a config file
            self.tags = dict(tuple(x.split("=")) for x in self.tags.split(","))

        self.mitiff_config = {}
        self.translate_channel_name = {}
        self.channel_order = {}
        self.palette = False
        self.sensor = None

    def save_image(self):
        """Save dataset as an image array."""
        raise NotImplementedError("save_image mitiff is not implemented.")

    def save_dataset(self, dataset, filename=None, fill_value=None,
                     compute=True, **kwargs):
        """Save single dataset as mitiff file."""
        LOG.debug("Starting in mitiff save_dataset ... ")

        def _delayed_create(dataset):
            try:
                if 'palette' in kwargs:
                    self.palette = kwargs['palette']
                _adjust_kwargs(dataset, kwargs)

                try:
                    self.mitiff_config[kwargs['sensor']] = dataset.attrs['metadata_requirements']['config']
                    self.channel_order[kwargs['sensor']] = dataset.attrs['metadata_requirements']['order']
                    self.file_pattern = dataset.attrs['metadata_requirements']['file_pattern']
                except KeyError:
                    # For some mitiff products this info is needed, for others not.
                    # If needed you should know how to fix this
                    pass

                try:
                    self.translate_channel_name[kwargs['sensor']] = \
                        dataset.attrs['metadata_requirements']['translate']
                except KeyError:
                    # For some mitiff products this info is needed, for others not.
                    # If needed you should know how to fix this
                    pass

                image_description = self._make_image_description(dataset, **kwargs)
                gen_filename = filename or self.get_filename(**dataset.attrs)
                LOG.info("Saving mitiff to: %s ...", gen_filename)
                self._save_datasets_as_mitiff(dataset, image_description,
                                              gen_filename, **kwargs)
            except (KeyError, ValueError, RuntimeError):
                raise

        delayed = dask.delayed(_delayed_create)(dataset)

        if compute:
            return delayed.compute()
        return delayed

    def save_datasets(self, datasets, filename=None, fill_value=None,
                      compute=True, **kwargs):
        """Save all datasets to one or more files."""
        LOG.debug("Starting in mitiff save_datasets ... ")

        def _delayed_create(datasets):
            dataset = datasets[0]

            try:
                _adjust_kwargs(dataset, kwargs)

                try:
                    self.mitiff_config[kwargs['sensor']] = dataset.attrs['metadata_requirements']['config']
                    translate = dataset.attrs['metadata_requirements']['translate']
                    self.translate_channel_name[kwargs['sensor']] = translate
                    self.channel_order[kwargs['sensor']] = dataset.attrs['metadata_requirements']['order']
                    self.file_pattern = dataset.attrs['metadata_requirements']['file_pattern']
                except KeyError:
                    # For some mitiff products this info is needed, for others not.
                    # If needed you should know how to fix this
                    pass

                image_description = self._make_image_description(datasets, **kwargs)
                LOG.debug("File pattern %s", self.file_pattern)
                if isinstance(datasets, list):
                    kwargs['start_time'] = dataset.attrs['start_time']
                else:
                    kwargs['start_time'] = datasets.attrs['start_time']
                gen_filename = filename or self.get_filename(**kwargs)
                LOG.info("Saving mitiff to: %s ...", gen_filename)
                self._save_datasets_as_mitiff(datasets, image_description, gen_filename, **kwargs)
            except (KeyError, ValueError, RuntimeError):
                raise

        delayed = dask.delayed(_delayed_create)(datasets)
        LOG.debug("About to call delayed compute ...")
        if compute:
            return delayed.compute()
        return delayed

    def _make_channel_list(self, datasets, **kwargs):
        channels = []
        try:
            if self.channel_order:
                channels = self._reorder_channels(datasets, **kwargs)
            elif self.palette:
                if 'palette_channel_name' in kwargs:
                    channels.append(kwargs['palette_channel_name'].upper())
                else:
                    LOG.error("Is palette but can not find palette_channel_name to name the dataset")
            else:
                for ch in range(len(datasets)):
                    channels.append(ch + 1)
        except KeyError:
            for ch in range(len(datasets)):
                channels.append(ch + 1)
        return channels

    def _reorder_channels(self, datasets, **kwargs):
        channels = []
        for cn in self.channel_order[kwargs['sensor']]:
            for ch, ds in enumerate(datasets):
                if isinstance(ds.attrs['prerequisites'][ch], (DataQuery, DataID)):
                    if ds.attrs['prerequisites'][ch]['name'] == cn:
                        channels.append(
                            ds.attrs['prerequisites'][ch]['name'])
                        break
                else:
                    if ds.attrs['prerequisites'][ch] == cn:
                        channels.append(
                            ds.attrs['prerequisites'][ch])
                        break
        return channels

    def _channel_names(self, channels, cns, **kwargs):
        _image_description = ""
        for ch in channels:
            try:
                _image_description += str(
                    self.mitiff_config[kwargs['sensor']][cns.get(ch, ch)]['alias'])
            except KeyError:
                _image_description += str(ch)
            _image_description += ' '

        # Replace last char(space) with \n
        _image_description = _image_description[:-1]
        _image_description += '\n'

        return _image_description

    def _add_sizes(self, datasets, first_dataset):
        _image_description = ' Xsize: '
        if isinstance(datasets, list):
            _image_description += str(first_dataset.sizes['x']) + '\n'
        else:
            _image_description += str(datasets.sizes['x']) + '\n'

        _image_description += ' Ysize: '
        if isinstance(datasets, list):
            _image_description += str(first_dataset.sizes['y']) + '\n'
        else:
            _image_description += str(datasets.sizes['y']) + '\n'

        return _image_description

    def _add_proj4_string(self, datasets, first_dataset):
        proj4_string = " Proj string: "

        if isinstance(datasets, list):
            area = first_dataset.attrs['area']
        else:
            area = datasets.attrs['area']
        # Use pyproj's CRS object to get a valid EPSG code if possible
        # only in newer pyresample versions with pyproj 2.0+ installed
        if hasattr(area, 'crs') and area.crs.to_epsg() is not None:
            proj4_string += "+init=EPSG:{}".format(area.crs.to_epsg())
        else:
            proj4_string += area.proj_str

        x_0 = 0
        y_0 = 0
        # FUTURE: Use pyproj 2.0+ to convert EPSG to PROJ4 if possible
        proj4_string, x_0 = self._convert_epsg_to_proj(proj4_string, x_0)

        if 'geos' in proj4_string:
            proj4_string = proj4_string.replace("+sweep=x ", "")
            if '+a=6378137.0 +b=6356752.31414' in proj4_string:
                proj4_string = proj4_string.replace("+a=6378137.0 +b=6356752.31414",
                                                    "+ellps=WGS84")
        if '+units=m' in proj4_string:
            proj4_string = proj4_string.replace("+units=m", "+units=km")

        if not any(datum in proj4_string for datum in ['datum', 'towgs84']):
            proj4_string += ' +towgs84=0,0,0'

        if 'units' not in proj4_string:
            proj4_string += ' +units=km'

        proj4_string = self._append_projection_center(proj4_string, datasets, first_dataset, x_0, y_0)
        LOG.debug("proj4_string: %s", proj4_string)
        proj4_string += '\n'

        return proj4_string

    def _append_projection_center(self, proj4_string, datasets, first_dataset, x_0, y_0):
        if isinstance(datasets, list):
            dataset = first_dataset
        else:
            dataset = datasets
        if 'x_0' not in proj4_string:
            proj4_string += ' +x_0=%.6f' % (
                    (-dataset.attrs['area'].area_extent[0] +
                     dataset.attrs['area'].pixel_size_x) + x_0)
            proj4_string += ' +y_0=%.6f' % (
                    (-dataset.attrs['area'].area_extent[1] +
                     dataset.attrs['area'].pixel_size_y) + y_0)
        elif '+x_0=0' in proj4_string and '+y_0=0' in proj4_string:
            proj4_string = proj4_string.replace("+x_0=0", '+x_0=%.6f' % (
                    (-dataset.attrs['area'].area_extent[0] +
                     dataset.attrs['area'].pixel_size_x) + x_0))
            proj4_string = proj4_string.replace("+y_0=0", '+y_0=%.6f' % (
                    (-dataset.attrs['area'].area_extent[1] +
                     dataset.attrs['area'].pixel_size_y) + y_0))
        return proj4_string

    def _convert_epsg_to_proj(self, proj4_string, x_0):
        if 'EPSG:32631' in proj4_string:
            proj4_string = proj4_string.replace("+init=EPSG:32631",
                                                "+proj=etmerc +lat_0=0 +lon_0=3 +k=0.9996 +ellps=WGS84 +datum=WGS84")
            x_0 = 500000
        elif 'EPSG:32632' in proj4_string:
            proj4_string = proj4_string.replace("+init=EPSG:32632",
                                                "+proj=etmerc +lat_0=0 +lon_0=9 +k=0.9996 +ellps=WGS84 +datum=WGS84")
            x_0 = 500000
        elif 'EPSG:32633' in proj4_string:
            proj4_string = proj4_string.replace("+init=EPSG:32633",
                                                "+proj=etmerc +lat_0=0 +lon_0=15 +k=0.9996 +ellps=WGS84 +datum=WGS84")
            x_0 = 500000
        elif 'EPSG:32634' in proj4_string:
            proj4_string = proj4_string.replace("+init=EPSG:32634",
                                                "+proj=etmerc +lat_0=0 +lon_0=21 +k=0.9996 +ellps=WGS84 +datum=WGS84")
            x_0 = 500000
        elif 'EPSG:32635' in proj4_string:
            proj4_string = proj4_string.replace("+init=EPSG:32635",
                                                "+proj=etmerc +lat_0=0 +lon_0=27 +k=0.9996 +ellps=WGS84 +datum=WGS84")
            x_0 = 500000
        elif 'EPSG' in proj4_string:
            LOG.warning("EPSG used in proj string but not converted. Please add this in code")
        return proj4_string, x_0

    def _add_pixel_sizes(self, datasets, first_dataset):
        _image_description = ""
        if isinstance(datasets, list):
            _image_description += ' Ax: %.6f' % (
                first_dataset.attrs['area'].pixel_size_x / 1000.)
            _image_description += ' Ay: %.6f' % (
                first_dataset.attrs['area'].pixel_size_y / 1000.)
        else:
            _image_description += ' Ax: %.6f' % (
                datasets.attrs['area'].pixel_size_x / 1000.)
            _image_description += ' Ay: %.6f' % (
                datasets.attrs['area'].pixel_size_y / 1000.)

        return _image_description

    def _add_corners(self, datasets, first_dataset):
        # But this ads up to upper left corner of upper left pixel.
        # But need to use the center of the pixel.
        # Therefor use the center of the upper left pixel.
        _image_description = ""
        if isinstance(datasets, list):
            _image_description += ' Bx: %.6f' % (
                first_dataset.attrs['area'].area_extent[0] / 1000. +
                first_dataset.attrs['area'].pixel_size_x / 1000. / 2.)  # LL_x
            _image_description += ' By: %.6f' % (
                first_dataset.attrs['area'].area_extent[3] / 1000. -
                first_dataset.attrs['area'].pixel_size_y / 1000. / 2.)  # UR_y
        else:
            _image_description += ' Bx: %.6f' % (
                datasets.attrs['area'].area_extent[0] / 1000. +
                datasets.attrs['area'].pixel_size_x / 1000. / 2.)  # LL_x
            _image_description += ' By: %.6f' % (
                datasets.attrs['area'].area_extent[3] / 1000. -
                datasets.attrs['area'].pixel_size_y / 1000. / 2.)  # UR_y

        _image_description += '\n'
        return _image_description

    def _add_calibration_datasets(self, ch, datasets, reverse_offset, reverse_scale, decimals):
        _reverse_offset = reverse_offset
        _reverse_scale = reverse_scale
        _decimals = decimals
        _table_calibration = ""
        found_calibration = False
        skip_calibration = False
        ds_list = datasets
        if not isinstance(datasets, list) and 'bands' not in datasets.sizes:
            ds_list = [datasets]

        for i, ds in enumerate(ds_list):
            if ('prerequisites' in ds.attrs and
                isinstance(ds.attrs['prerequisites'], list) and
                len(ds.attrs['prerequisites']) >= i + 1 and
                    isinstance(ds.attrs['prerequisites'][i], (DataQuery, DataID))):
                if ds.attrs['prerequisites'][i].get('name') == str(ch):
                    if ds.attrs['prerequisites'][i].get('calibration') == 'RADIANCE':
                        raise NotImplementedError(
                            "Mitiff radiance calibration not implemented.")
                    # _table_calibration += ', Radiance, '
                    # _table_calibration += '[W/m²/µm/sr]'
                    # _decimals = 8
                    elif ds.attrs['prerequisites'][i].get('calibration') == 'brightness_temperature':
                        found_calibration = True
                        _table_calibration += ', BT, '
                        _table_calibration += "\N{DEGREE SIGN}"
                        _table_calibration += u'[C]'

                        _reverse_offset = 255.
                        _reverse_scale = -1.
                        _decimals = 2
                    elif ds.attrs['prerequisites'][i].get('calibration') == 'reflectance':
                        found_calibration = True
                        _table_calibration += ', Reflectance(Albedo), '
                        _table_calibration += '[%]'
                        _decimals = 2
                    else:
                        LOG.warning("Unknown calib type. Must be Radiance, Reflectance or BT.")

                        break
                else:
                    continue
            else:
                _table_calibration = ""
                skip_calibration = True
                break
        if not found_calibration:
            _table_calibration = ""
            skip_calibration = True
            # How to format string by passing the format
            # http://stackoverflow.com/questions/1598579/rounding-decimals-with-new-python-format-function
        return skip_calibration, _table_calibration, _reverse_offset, _reverse_scale, _decimals

    def _add_palette_info(self, datasets, palette_unit, palette_description, **kwargs):
        # mitiff key word for palette interpretion
        _palette = '\n COLOR INFO:\n'
        # mitiff info for the unit of the interpretion
        _palette += ' {}\n'.format(palette_unit)
        # The length of the palette description as needed by mitiff in DIANA
        _palette += ' {}\n'.format(len(palette_description))
        for desc in palette_description:
            _palette += ' {}\n'.format(desc)
        return _palette

    def _add_calibration(self, channels, cns, datasets, **kwargs):

        _table_calibration = ""
        skip_calibration = False
        for ch in channels:

            palette = False
            # Make calibration.
            if palette:
                raise NotImplementedError("Mitiff palette saving is not implemented.")
            else:
                _table_calibration += 'Table_calibration: '
                try:
                    _table_calibration += str(
                        self.mitiff_config[kwargs['sensor']][cns.get(ch, ch)]['alias'])
                except KeyError:
                    _table_calibration += str(ch)

                _reverse_offset = 0.
                _reverse_scale = 1.
                _decimals = 2

                skip_calibration, __table_calibration, _reverse_offset, _reverse_scale, _decimals = \
                    self._add_calibration_datasets(ch, datasets, _reverse_offset, _reverse_scale, _decimals)
                _table_calibration += __table_calibration

                if not skip_calibration:
                    _table_calibration += ', 8, [ '
                    for val in range(0, 256):
                        # Comma separated list of values
                        _table_calibration += '{0:.{1}f} '.format((float(self.mitiff_config[
                            kwargs['sensor']][cns.get(ch, ch)]['min-val']) +
                            ((_reverse_offset + _reverse_scale * val) *
                             (float(self.mitiff_config[kwargs['sensor']][cns.get(ch, ch)]['max-val']) -
                             float(self.mitiff_config[kwargs['sensor']][cns.get(ch, ch)]['min-val']))) / 255.),
                            _decimals)
                        # _table_calibration += '0.00000000 '

                    _table_calibration += ']\n\n'
                else:
                    _table_calibration = ""

        return _table_calibration

    def _make_image_description(self, datasets, **kwargs):
        r"""Generate image description for mitiff.

        Satellite: NOAA 18
        Date and Time: 06:58 31/05-2016
        SatDir: 0
        Channels:   6 In this file: 1-VIS0.63 2-VIS0.86 3(3B)-IR3.7
        4-IR10.8 5-IR11.5 6(3A)-VIS1.6
        Xsize:  4720
        Ysize:  5544
        Map projection: Stereographic
        Proj string: +proj=stere +lon_0=0 +lat_0=90 +lat_ts=60
        +ellps=WGS84 +towgs84=0,0,0 +units=km
        +x_0=2526000.000000 +y_0=5806000.000000
        TrueLat: 60 N
        GridRot: 0
        Xunit:1000 m Yunit: 1000 m
        NPX: 0.000000 NPY: 0.000000
        Ax: 1.000000 Ay: 1.000000 Bx: -2526.000000 By: -262.000000

        Satellite: <satellite name>
        Date and Time: <HH:MM dd/mm-yyyy>
        SatDir: 0
        Channels:   <number of chanels> In this file: <channels names in order>
        Xsize:  <number of pixels x>
        Ysize:  <number of pixels y>
        Map projection: Stereographic
        Proj string: <proj4 string with +x_0 and +y_0 which is the positive
        distance from proj origo
        to the lower left corner of the image data>
        TrueLat: 60 N
        GridRot: 0
        Xunit:1000 m Yunit: 1000 m
        NPX: 0.000000 NPY: 0.000000
        Ax: <pixels size x in km> Ay: <pixel size y in km> Bx: <left corner of
        upper right pixel in km>
        By: <upper corner of upper right pixel in km>


        if palette image write special palette
        if normal channel write table calibration:
        Table_calibration: <channel name>, <calibration type>, [<unit>],
        <no of bits of data>,
        [<calibration values space separated>]\n\n

        """
        translate_platform_name = {'metop01': 'Metop-B',
                                   'metop02': 'Metop-A',
                                   'metop03': 'Metop-C',
                                   'noaa15': 'NOAA-15',
                                   'noaa16': 'NOAA-16',
                                   'noaa17': 'NOAA-17',
                                   'noaa18': 'NOAA-18',
                                   'noaa19': 'NOAA-19'}

        first_dataset = datasets
        if isinstance(datasets, list):
            LOG.debug("Datasets is a list of dataset")
            first_dataset = datasets[0]

        _platform_name = self._get_platform_name(first_dataset, translate_platform_name, kwargs)

        _image_description = ''
        _image_description.encode('utf-8')

        _image_description += ' Satellite: '
        if _platform_name is not None:
            _image_description += _platform_name

        _image_description += '\n'

        _image_description += ' Date and Time: '
        # Select earliest start_time
        first = True
        earliest = 0
        for dataset in datasets:
            if first:
                earliest = dataset.attrs['start_time']
            else:
                if dataset.attrs['start_time'] < earliest:
                    earliest = dataset.attrs['start_time']
            first = False
        LOG.debug("earliest start_time: %s", earliest)
        _image_description += earliest.strftime("%H:%M %d/%m-%Y\n")

        _image_description += ' SatDir: 0\n'

        _image_description += ' Channels: '

        _image_description += self._get_dataset_len(datasets)
        _image_description += ' In this file: '

        channels = self._make_channel_list(datasets, **kwargs)

        try:
            cns = self.translate_channel_name.get(kwargs['sensor'], {})
        except KeyError:
            pass

        _image_description += self._channel_names(channels, cns, **kwargs)

        _image_description += self._add_sizes(datasets, first_dataset)

        _image_description += ' Map projection: Stereographic\n'

        _image_description += self._add_proj4_string(datasets, first_dataset)

        _image_description += ' TrueLat: 60N\n'
        _image_description += ' GridRot: 0\n'

        _image_description += ' Xunit:1000 m Yunit: 1000 m\n'

        _image_description += ' NPX: %.6f' % (0)
        _image_description += ' NPY: %.6f' % (0) + '\n'

        _image_description += self._add_pixel_sizes(datasets, first_dataset)
        _image_description += self._add_corners(datasets, first_dataset)

        if isinstance(datasets, list):
            LOG.debug("Area extent: %s", first_dataset.attrs['area'].area_extent)
        else:
            LOG.debug("Area extent: %s", datasets.attrs['area'].area_extent)

        if self.palette:
            LOG.debug("Doing palette image")
            _image_description += self._add_palette_info(datasets, **kwargs)
        else:
            _image_description += self._add_calibration(channels, cns, datasets, **kwargs)

        return _image_description

    def _get_dataset_len(self, datasets):
        if isinstance(datasets, list):
            LOG.debug("len datasets: %s", len(datasets))
            dataset_len = str(len(datasets))
        elif 'bands' in datasets.sizes:
            LOG.debug("len datasets: %s", datasets.sizes['bands'])
            dataset_len = str(datasets.sizes['bands'])
        elif len(datasets.sizes) == 2:
            LOG.debug("len datasets: 1")
            dataset_len = '1'
        else:
            dataset_len = ""
        return dataset_len

    def _get_platform_name(self, first_dataset, translate_platform_name, kwargs):
        if 'platform_name' in first_dataset.attrs:
            _platform_name = translate_platform_name.get(
                first_dataset.attrs['platform_name'],
                first_dataset.attrs['platform_name'])
        elif 'platform_name' in kwargs:
            _platform_name = translate_platform_name.get(
                kwargs['platform_name'], kwargs['platform_name'])
        else:
            _platform_name = None
        return _platform_name

    def _calibrate_data(self, dataset, calibration, min_val, max_val):
        reverse_offset = 0.
        reverse_scale = 1.
        if calibration == 'brightness_temperature':
            # If data is brightness temperature, the data must be inverted.
            reverse_offset = 255.
            reverse_scale = -1.
            dataset.data += KELVIN_TO_CELSIUS

        # Need to possible translate channels names from satpy to mitiff
        _data = reverse_offset + reverse_scale * ((dataset.data - float(min_val)) /
                                                  (float(max_val) - float(min_val))) * 255.
        return _data.clip(0, 255)

    def _save_as_palette(self, datasets, tmp_gen_filename, tiffinfo, **kwargs):
        # MITIFF palette has only one data channel
        if len(datasets.dims) == 2:
            LOG.debug("Palette ok with only 2 dimensions. ie only x and y")
            # 3 = Palette color. In this model, a color is described with a single component.
            # The value of the component is used as an index into the red, green and blue curves
            # in the ColorMap field to retrieve an RGB triplet that defines the color. When
            # PhotometricInterpretation=3 is used, ColorMap must be present and SamplesPerPixel must be 1.
            tiffinfo[270] = tiffinfo[270].decode('utf-8')

            img = Image.fromarray(datasets.data.astype(np.uint8), mode='P')
            if 'palette_color_map' in kwargs:
                img.putpalette(ImagePalette.ImagePalette('RGB', kwargs['palette_color_map']))
            else:
                LOG.error("In a mitiff palette image a color map must be provided: palette_color_map is missing.")
                return

            img.save(tmp_gen_filename, compression='tiff_deflate', compress_level=9, tiffinfo=tiffinfo)

    def _save_as_enhanced(self, datasets, tmp_gen_filename, **kwargs):
        """Save datasets as an enhanced RGB image."""
        img = get_enhanced_image(datasets.squeeze(), enhance=self.enhancer)
        tiffinfo = {}
        if 'bands' in img.data.sizes and 'bands' not in datasets.sizes:
            LOG.debug("Datasets without 'bands' become image with 'bands' due to enhancement.")
            LOG.debug("Needs to regenerate mitiff image description")
        image_description = self._make_image_description(img.data, **kwargs)
        tiffinfo[IMAGEDESCRIPTION] = (image_description).encode('utf-8')

        mitiff_frames = []
        for band in img.data['bands']:
            chn = img.data.sel(bands=band)
            data = chn.values.clip(0, 1) * 254. + 1
            data = data.clip(0, 255)
            mitiff_frames.append(Image.fromarray(data.astype(np.uint8), mode='L'))
        mitiff_frames[0].save(tmp_gen_filename, save_all=True, append_images=mitiff_frames[1:],
                              compression='tiff_deflate', compress_level=9, tiffinfo=tiffinfo)

    def _generate_intermediate_filename(self, gen_filename):
        """Replace mitiff ext because pillow doesn't recognise the file type."""
        bs, ex = os.path.splitext(gen_filename)
        tmp_gen_filename = gen_filename
        if ex.endswith('mitiff'):
            bd = os.path.dirname(bs)
            bn = os.path.basename(bs)
            tmp_gen_filename = os.path.join(bd, '.' + bn + '.tif')
        return tmp_gen_filename

    def _save_datasets_as_mitiff(self, datasets, image_description,
                                 gen_filename, **kwargs):
        """Put all together and save as a tiff file.

        Include the special tags making it a mitiff file.

        """
        tmp_gen_filename = self._generate_intermediate_filename(gen_filename)
        tiffinfo = {}
        tiffinfo[IMAGEDESCRIPTION] = (image_description).encode('latin-1')

        cns = self.translate_channel_name.get(kwargs['sensor'], {})
        if isinstance(datasets, list):
            LOG.debug("Saving datasets as list")
            mitiff_frames = []
            for _cn in self.channel_order[kwargs['sensor']]:
                for dataset in datasets:
                    if dataset.attrs['name'] == _cn:
                        # Need to possible translate channels names from satpy to mitiff
                        cn = cns.get(dataset.attrs['name'], dataset.attrs['name'])
                        data = self._calibrate_data(dataset, dataset.attrs['calibration'],
                                                    self.mitiff_config[kwargs['sensor']][cn]['min-val'],
                                                    self.mitiff_config[kwargs['sensor']][cn]['max-val'])
                        mitiff_frames.append(Image.fromarray(data.astype(np.uint8), mode='L'))
                        break
            mitiff_frames[0].save(tmp_gen_filename, save_all=True, append_images=mitiff_frames[1:],
                                  compression='tiff_deflate', compress_level=9, tiffinfo=tiffinfo)
        elif 'dataset' in datasets.attrs['name']:
            LOG.debug("Saving dataset as single dataset.")
            self._save_single_dataset(datasets, cns, tmp_gen_filename, tiffinfo, kwargs)
        elif self.palette:
            LOG.debug("Saving dataset as palette.")
            self._save_as_palette(datasets, tmp_gen_filename, tiffinfo, **kwargs)
        else:
            LOG.debug("Saving datasets as enhanced image")
            self._save_as_enhanced(datasets, tmp_gen_filename, **kwargs)
        os.rename(tmp_gen_filename, gen_filename)

    def _save_single_dataset(self, datasets, cns, tmp_gen_filename, tiffinfo, kwargs):
        LOG.debug("Saving %s as a dataset.", datasets.attrs['name'])
        if len(datasets.dims) == 2 and (all('bands' not in i for i in datasets.dims)):
            # Special case with only one channel ie. no bands

            # Need to possible translate channels names from satpy to mitiff
            # Note the last index is a tuple index.
            cn = cns.get(datasets.attrs['prerequisites'][0]['name'],
                         datasets.attrs['prerequisites'][0]['name'])
            data = self._calibrate_data(datasets, datasets.attrs['prerequisites'][0].get('calibration'),
                                        self.mitiff_config[kwargs['sensor']][cn]['min-val'],
                                        self.mitiff_config[kwargs['sensor']][cn]['max-val'])
            Image.fromarray(data.astype(np.uint8)).save(tmp_gen_filename, compression='tiff_deflate',
                                                        compress_level=9, tiffinfo=tiffinfo)
        else:
            for _cn_i, _cn in enumerate(self.channel_order[kwargs['sensor']]):
                for band in datasets['bands']:
                    if band == _cn:
                        chn = datasets.sel(bands=band)
                        # Need to possible translate channels names from satpy to mitiff
                        # Note the last index is a tuple index.
                        cn = cns.get(chn.attrs['prerequisites'][_cn_i]['name'],
                                     chn.attrs['prerequisites'][_cn_i]['name'])
                        data = self._calibrate_data(chn, chn.attrs['prerequisites'][_cn_i].get('calibration'),
                                                    self.mitiff_config[kwargs['sensor']][cn]['min-val'],
                                                    self.mitiff_config[kwargs['sensor']][cn]['max-val'])

                        Image.fromarray(data.astype(np.uint8)).save(tmp_gen_filename, compression='tiff_deflate',
                                                                    compress_level=9, tiffinfo=tiffinfo)
                        break