Source code for satpy.enhancements

#!/usr/bin/env python
# Copyright (c) 2017 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/>.
"""Enhancements."""

import logging
import os
import warnings
from collections import namedtuple
from functools import wraps
from numbers import Number
from typing import Optional

import dask
import dask.array as da
import numpy as np
import xarray as xr
from trollimage.xrimage import XRImage

from satpy._compat import ArrayLike
from satpy._config import get_config_path

LOG = logging.getLogger(__name__)


[docs]def stretch(img, **kwargs): """Perform stretch.""" return img.stretch(**kwargs)
[docs]def gamma(img, **kwargs): """Perform gamma correction.""" return img.gamma(**kwargs)
[docs]def invert(img, *args): """Perform inversion.""" return img.invert(*args)
[docs]def exclude_alpha(func): """Exclude the alpha channel from the DataArray before further processing.""" @wraps(func) def wrapper(data, **kwargs): bands = data.coords['bands'].values exclude = ['A'] if 'A' in bands else [] band_data = data.sel(bands=[b for b in bands if b not in exclude]) band_data = func(band_data, **kwargs) attrs = data.attrs attrs.update(band_data.attrs) # combine the new data with the excluded data new_data = xr.concat([band_data, data.sel(bands=exclude)], dim='bands') data.data = new_data.sel(bands=bands).data data.attrs = attrs return data return wrapper
[docs]def on_separate_bands(func): """Apply `func` one band of the DataArray at a time. If this decorator is to be applied along with `on_dask_array`, this decorator has to be applied first, eg:: @on_separate_bands @on_dask_array def my_enhancement_function(data): ... """ @wraps(func) def wrapper(data, **kwargs): attrs = data.attrs data_arrs = [] for idx, band in enumerate(data.coords['bands'].values): band_data = func(data.sel(bands=[band]), index=idx, **kwargs) data_arrs.append(band_data) # we assume that the func can add attrs attrs.update(band_data.attrs) data.data = xr.concat(data_arrs, dim='bands').data data.attrs = attrs return data return wrapper
[docs]def on_dask_array(func): """Pass the underlying dask array to *func* instead of the xarray.DataArray.""" @wraps(func) def wrapper(data, **kwargs): dims = data.dims coords = data.coords d_arr = func(data.data, **kwargs) return xr.DataArray(d_arr, dims=dims, coords=coords) return wrapper
[docs]def using_map_blocks(func): """Run the provided function using :func:`dask.array.core.map_blocks`. This means dask will call the provided function with a single chunk as a numpy array. """ @wraps(func) def wrapper(data, **kwargs): return da.map_blocks(func, data, meta=np.array((), dtype=data.dtype), dtype=data.dtype, chunks=data.chunks, **kwargs) return on_dask_array(wrapper)
[docs]def crefl_scaling(img, **kwargs): """Apply non-linear stretch used by CREFL-based RGBs.""" LOG.debug("Applying the crefl_scaling") warnings.warn("'crefl_scaling' is deprecated, use 'piecewise_linear_stretch' instead.", DeprecationWarning) img.data.data = img.data.data / 100 return piecewise_linear_stretch(img, xp=kwargs['idx'], fp=kwargs['sc'], reference_scale_factor=255)
[docs]def piecewise_linear_stretch( img: XRImage, xp: ArrayLike, fp: ArrayLike, reference_scale_factor: Optional[Number] = None, **kwargs) -> xr.DataArray: """Apply 1D linear interpolation. This uses :func:`numpy.interp` mapped over the provided dask array chunks. Args: img: Image data to be scaled. It is assumed the data is already normalized between 0 and 1. xp: Input reference values of the image data points used for interpolation. This is passed directly to :func:`numpy.interp`. fp: Target reference values of the output image data points used for interpolation. This is passed directly to :func:`numpy.interp`. reference_scale_factor: Divide ``xp`` and ``fp`` by this value before using them for interpolation. This is a convenience to make matching normalized image data to interp coordinates or to avoid floating point precision errors in YAML configuration files. If not provided, ``xp`` and ``fp`` will not be modified. Examples: This example YAML uses a 'crude' stretch to pre-scale the RGB data and then uses reference points in a 0-255 range. .. code-block:: yaml true_color_linear_interpolation: sensor: abi standard_name: true_color operations: - name: reflectance_range method: !!python/name:satpy.enhancements.stretch kwargs: {stretch: 'crude', min_stretch: 0., max_stretch: 100.} - name: Linear interpolation method: !!python/name:satpy.enhancements.piecewise_linear_stretch kwargs: xp: [0., 25., 55., 100., 255.] fp: [0., 90., 140., 175., 255.] reference_scale_factor: 255 This example YAML does the same as the above on the C02 channel, but the interpolation reference points are already adjusted for the input reflectance (%) data and the output range (0 to 1). .. code-block:: yaml c02_linear_interpolation: sensor: abi standard_name: C02 operations: - name: Linear interpolation method: !!python/name:satpy.enhancements.piecewise_linear_stretch kwargs: xp: [0., 9.8039, 21.5686, 39.2157, 100.] fp: [0., 0.3529, 0.5490, 0.6863, 1.0] """ LOG.debug("Applying the piecewise_linear_stretch") if reference_scale_factor is not None: xp = np.asarray(xp) / reference_scale_factor fp = np.asarray(fp) / reference_scale_factor return _piecewise_linear(img.data, xp=xp, fp=fp)
@exclude_alpha @using_map_blocks def _piecewise_linear(band_data, xp, fp): # Interpolate band on [0,1] using "lazy" arrays (put calculations off until the end). interp_data = np.interp(band_data, xp=xp, fp=fp) interp_data = np.clip(interp_data, 0, 1, out=interp_data) return interp_data
[docs]def cira_stretch(img, **kwargs): """Logarithmic stretch adapted to human vision. Applicable only for visible channels. """ LOG.debug("Applying the cira-stretch") return _cira_stretch(img.data)
@exclude_alpha def _cira_stretch(band_data): log_root = np.log10(0.0223) denom = (1.0 - log_root) * 0.75 band_data *= 0.01 band_data = band_data.clip(np.finfo(float).eps) band_data = np.log10(band_data) band_data -= log_root band_data /= denom return band_data
[docs]def reinhard_to_srgb(img, saturation=1.25, white=100, **kwargs): """Stretch method based on the Reinhard algorithm, using luminance. Args: saturation: Saturation enhancement factor. Less is grayer. Neutral is 1. white: the reflectance luminance to set to white (in %). Reinhard, Erik & Stark, Michael & Shirley, Peter & Ferwerda, James. (2002). Photographic Tone Reproduction For Digital Images. ACM Transactions on Graphics. :doi: `21. 10.1145/566654.566575` """ with xr.set_options(keep_attrs=True): # scale the data to [0, 1] interval rgb = img.data / 100 white /= 100 # extract color components r = rgb.sel(bands='R').data g = rgb.sel(bands='G').data b = rgb.sel(bands='B').data # saturate luma = _compute_luminance_from_rgb(r, g, b) rgb = (luma + (rgb - luma) * saturation).clip(0) # reinhard reinhard_luma = (luma / (1 + luma)) * (1 + luma/(white**2)) coef = reinhard_luma / luma rgb = rgb * coef # srgb gamma rgb.data = _srgb_gamma(rgb.data) img.data = rgb return img.data
def _compute_luminance_from_rgb(r, g, b): """Compute the luminance of the image.""" return r * 0.2126 + g * 0.7152 + b * 0.0722 def _srgb_gamma(arr): """Apply the srgb gamma.""" return da.where(arr < 0.0031308, arr * 12.92, 1.055 * arr ** 0.41666 - 0.055)
[docs]def lookup(img, **kwargs): """Assign values to channels based on a table.""" luts = np.array(kwargs['luts'], dtype=np.float32) / 255.0 return _lookup_table(img.data, luts=luts)
@exclude_alpha @on_separate_bands @using_map_blocks def _lookup_table(band_data, luts=None, index=-1): # NaN/null values will become 0 lut = luts[:, index] if len(luts.shape) == 2 else luts band_data = band_data.clip(0, lut.size - 1).astype(np.uint8) return lut[band_data]
[docs]def colorize(img, **kwargs): """Colorize the given image. Args: img: image to be colorized Kwargs: palettes: colormap(s) to use The `palettes` kwarg can be one of the following: - a trollimage.colormap.Colormap object - list of dictionaries with each of one of the following forms: - {'filename': '/path/to/colors.npy', 'min_value': <float, min value to match colors to>, 'max_value': <float, min value to match colors to>, 'reverse': <bool, reverse the colormap if True (default: False)} - {'colors': <trollimage.colormap.Colormap instance>, 'min_value': <float, min value to match colors to>, 'max_value': <float, min value to match colors to>, 'reverse': <bool, reverse the colormap if True (default: False)} - {'colors': <tuple of RGB(A) tuples>, 'min_value': <float, min value to match colors to>, 'max_value': <float, min value to match colors to>, 'reverse': <bool, reverse the colormap if True (default: False)} - {'colors': <tuple of RGB(A) tuples>, 'values': <tuple of values to match colors to>, 'min_value': <float, min value to match colors to>, 'max_value': <float, min value to match colors to>, 'reverse': <bool, reverse the colormap if True (default: False)} If multiple palettes are supplied, they are concatenated before applied. """ full_cmap = _merge_colormaps(kwargs) img.colorize(full_cmap)
[docs]def palettize(img, **kwargs): """Palettize the given image (no color interpolation).""" full_cmap = _merge_colormaps(kwargs) img.palettize(full_cmap)
def _merge_colormaps(kwargs): """Merge colormaps listed in kwargs.""" from trollimage.colormap import Colormap full_cmap = None palette = kwargs['palettes'] if isinstance(palette, Colormap): full_cmap = palette else: for itm in palette: cmap = create_colormap(itm) if full_cmap is None: full_cmap = cmap else: full_cmap = full_cmap + cmap return full_cmap
[docs]def create_colormap(palette): """Create colormap of the given numpy file, color vector, or colormap. Args: palette (dict): Information describing how to create a colormap object. See below for more details. **From a file** Colormaps can be loaded from ``.npy``, ``.npz``, or comma-separated text files. Numpy (npy/npz) files should be 2D arrays with rows for each color. Comma-separated files should have a row for each color with each column representing a single value/channel. The filename to load can be provided with the ``filename`` key in the provided palette information. A filename ending with ``.npy`` or ``.npz`` is read as a numpy file with :func:`numpy.load`. All other extensions are read as a comma-separated file. For ``.npz`` files the data must be stored as a positional list where the first element represents the colormap to use. See :func:`numpy.savez` for more information. The path to the colormap can be relative if it is stored in a directory specified by :ref:`config_path_setting`. Otherwise it should be an absolute path. The colormap is interpreted as 1 of 4 different "colormap modes": ``RGB``, ``RGBA``, ``VRGB``, or ``VRGBA``. The colormap mode can be forced with the ``colormap_mode`` key in the provided palette information. If it is not provided then a default will be chosen based on the number of columns in the array (3: RGB, 4: VRGB, 5: VRGBA). The "V" in the possible colormap modes represents the control value of where that color should be applied. If "V" is not provided in the colormap data it defaults to the row index in the colormap array (0, 1, 2, ...) divided by the total number of colors to produce a number between 0 and 1. See the "Set Range" section below for more information. The remaining elements in the colormap array represent the Red (R), Green (G), and Blue (B) color to be mapped to. See the "Color Scale" section below for more information on the value range of provided numbers. **From a list** Colormaps can be loaded from lists of colors provided by the ``colors`` key in the provided dictionary. Each element in the list represents a single color to be mapped to and can be 3 (RGB) or 4 (RGBA) elements long. By default the value or control point for a color is determined by the index in the list (0, 1, 2, ...) divided by the total number of colors to produce a number between 0 and 1. This can be overridden by providing a ``values`` key in the provided dictionary. See the "Set Range" section below for more information. See the "Color Scale" section below for more information on the value range of provided numbers. **From a builtin colormap** Colormaps can be loaded by name from the builtin colormaps in the ``trollimage``` package. Specify the name with the ``colors`` key in the provided dictionary (ex. ``{'colors': 'blues'}``). See :doc:`trollimage:colormap` for the full list of available colormaps. **Color Scale** By default colors are expected to be in a 0-255 range. This can be overridden by specifying ``color_scale`` in the provided colormap information. A common alternative to 255 is ``1`` to specify floating point numbers between 0 and 1. The resulting Colormap uses the normalized color values (0-1). **Set Range** By default the control points or values of the Colormap are between 0 and 1. This means that data values being mapped to a color must also be between 0 and 1. When this is not the case, the expected input range of the data can be used to configure the Colormap and change the control point values. To do this specify the input data range with ``min_value`` and ``max_value``. See :meth:`trollimage.colormap.Colormap.set_range` for more information. """ fname = palette.get('filename', None) colors = palette.get('colors', None) # are colors between 0-255 or 0-1 color_scale = palette.get('color_scale', 255) if fname: cmap = _create_colormap_from_file(fname, palette, color_scale) elif isinstance(colors, (tuple, list)): cmap = _create_colormap_from_sequence(colors, palette, color_scale) elif isinstance(colors, str): import copy from trollimage import colormap cmap = copy.copy(getattr(colormap, colors)) else: raise ValueError("Unknown colormap format: {}".format(palette)) if palette.get("reverse", False): cmap.reverse() if 'min_value' in palette and 'max_value' in palette: cmap.set_range(palette["min_value"], palette["max_value"]) elif 'min_value' in palette or 'max_value' in palette: raise ValueError("Both 'min_value' and 'max_value' must be specified") return cmap
def _create_colormap_from_sequence(colors, palette, color_scale): from trollimage.colormap import Colormap cmap = [] values = palette.get('values', None) for idx, color in enumerate(colors): if values is not None: value = values[idx] else: value = idx / float(len(colors) - 1) if color_scale != 1: color = tuple(elem / float(color_scale) for elem in color) cmap.append((value, tuple(color))) return Colormap(*cmap) def _create_colormap_from_file(filename, palette, color_scale): from trollimage.colormap import Colormap data = _read_colormap_data_from_file(filename) cols = data.shape[1] default_modes = { 3: 'RGB', 4: 'VRGB', 5: 'VRGBA' } default_mode = default_modes.get(cols) mode = palette.setdefault('colormap_mode', default_mode) if mode is None or len(mode) != cols: raise ValueError( "Unexpected colormap shape for mode '{}'".format(mode)) rows = data.shape[0] if mode[0] == 'V': colors = data[:, 1:] if color_scale != 1: colors = data[:, 1:] / float(color_scale) values = data[:, 0] else: colors = data if color_scale != 1: colors = colors / float(color_scale) values = np.arange(rows) / float(rows - 1) return Colormap(*zip(values, colors)) def _read_colormap_data_from_file(filename): if not os.path.exists(filename): filename = get_config_path(filename) ext = os.path.splitext(filename)[1] if ext in (".npy", ".npz"): file_content = np.load(filename) if ext == ".npz": # .npz is a collection # assume position list-like and get the first element file_content = file_content["arr_0"] return file_content # CSV return np.loadtxt(filename, delimiter=",")
[docs]def three_d_effect(img, **kwargs): """Create 3D effect using convolution.""" w = kwargs.get('weight', 1) LOG.debug("Applying 3D effect with weight %.2f", w) kernel = np.array([[-w, 0, w], [-w, 1, w], [-w, 0, w]]) mode = kwargs.get('convolve_mode', 'same') return _three_d_effect(img.data, kernel=kernel, mode=mode)
@exclude_alpha @on_separate_bands @on_dask_array def _three_d_effect(band_data, kernel=None, mode=None, index=None): del index delay = dask.delayed(_three_d_effect_delayed)(band_data, kernel, mode) new_data = da.from_delayed(delay, shape=band_data.shape, dtype=band_data.dtype) return new_data def _three_d_effect_delayed(band_data, kernel, mode): """Kernel for running delayed 3D effect creation.""" from scipy.signal import convolve2d band_data = band_data.reshape(band_data.shape[1:]) new_data = convolve2d(band_data, kernel, mode=mode) return new_data.reshape((1, band_data.shape[0], band_data.shape[1]))
[docs]def btemp_threshold(img, min_in, max_in, threshold, threshold_out=None, **kwargs): """Scale data linearly in two separate regions. This enhancement scales the input data linearly by splitting the data into two regions; min_in to threshold and threshold to max_in. These regions are mapped to 1 to threshold_out and threshold_out to 0 respectively, resulting in the data being "flipped" around the threshold. A default threshold_out is set to `176.0 / 255.0` to match the behavior of the US National Weather Service's forecasting tool called AWIPS. Args: img (XRImage): Image object to be scaled min_in (float): Minimum input value to scale max_in (float): Maximum input value to scale threshold (float): Input value where to split data in to two regions threshold_out (float): Output value to map the input `threshold` to. Optional, defaults to 176.0 / 255.0. """ threshold_out = threshold_out if threshold_out is not None else (176 / 255.0) low_factor = (threshold_out - 1.) / (min_in - threshold) low_offset = 1. + (low_factor * min_in) high_factor = threshold_out / (max_in - threshold) high_offset = high_factor * max_in Coeffs = namedtuple("Coeffs", "factor offset") high = Coeffs(high_factor, high_offset) low = Coeffs(low_factor, low_offset) return _bt_threshold(img.data, threshold=threshold, high_coeffs=high, low_coeffs=low)
@exclude_alpha @using_map_blocks def _bt_threshold(band_data, threshold, high_coeffs, low_coeffs): # expects dask array to be passed return np.where(band_data >= threshold, high_coeffs.offset - high_coeffs.factor * band_data, low_coeffs.offset - low_coeffs.factor * band_data)