def __call__(self, src_ds):
logger.info("Applying ReprojectionOptimization")
# setup
src_sr = osr.SpatialReference()
src_sr.ImportFromWkt(src_ds.GetProjection())
dst_sr = osr.SpatialReference()
dst_sr.ImportFromEPSG(self.srid)
if src_sr.IsSame(dst_sr) and (src_ds.GetGeoTransform()[1] > 0) \
and (src_ds.GetGeoTransform()[5] < 0) \
and (src_ds.GetGeoTransform()[2] == 0) \
and (src_ds.GetGeoTransform()[4] == 0):
logger.info(
"Source and destination projection are equal and image "
"is not flipped or has rotated axes. Thus, no "
"reprojection is required.")
return src_ds
# create a temporary dataset to get information about the output size
tmp_ds = gdal.AutoCreateWarpedVRT(src_ds, None, dst_sr.ExportToWkt(),
gdal.GRA_Bilinear, 0.125)
try:
# create the output dataset
dst_ds = create_temp(tmp_ds.RasterXSize,
tmp_ds.RasterYSize,
src_ds.RasterCount,
src_ds.GetRasterBand(1).DataType,
temp_root=self.temporary_directory)
# initialize with no data
for i in range(src_ds.RasterCount):
src_band = src_ds.GetRasterBand(i + 1)
if src_band.GetNoDataValue() is not None:
dst_band = dst_ds.GetRasterBand(i + 1)
dst_band.SetNoDataValue(src_band.GetNoDataValue())
dst_band.Fill(src_band.GetNoDataValue())
# reproject the image
dst_ds.SetProjection(dst_sr.ExportToWkt())
dst_ds.SetGeoTransform(tmp_ds.GetGeoTransform())
gdal.ReprojectImage(src_ds, dst_ds, src_sr.ExportToWkt(),
dst_sr.ExportToWkt(), gdal.GRA_Bilinear)
tmp_ds = None
# copy the metadata
copy_metadata(src_ds, dst_ds)
return dst_ds
except:
cleanup_temp(dst_ds)
raise
def __call__(self, src_ds):
logger.info("Applying ReprojectionOptimization")
# setup
src_sr = osr.SpatialReference()
src_sr.ImportFromWkt(src_ds.GetProjection())
dst_sr = osr.SpatialReference()
dst_sr.ImportFromEPSG(self.srid)
if src_sr.IsSame(dst_sr) and (src_ds.GetGeoTransform()[1] > 0) \
and (src_ds.GetGeoTransform()[5] < 0) \
and (src_ds.GetGeoTransform()[2] == 0) \
and (src_ds.GetGeoTransform()[4] == 0):
logger.info("Source and destination projection are equal and image "
"is not flipped or has rotated axes. Thus, no "
"reprojection is required.")
return src_ds
# create a temporary dataset to get information about the output size
tmp_ds = gdal.AutoCreateWarpedVRT(src_ds, None, dst_sr.ExportToWkt(),
gdal.GRA_Bilinear, 0.125)
try:
# create the output dataset
dst_ds = create_temp(tmp_ds.RasterXSize, tmp_ds.RasterYSize,
src_ds.RasterCount,
src_ds.GetRasterBand(1).DataType,
temp_root=self.temporary_directory)
# initialize with no data
for i in range(src_ds.RasterCount):
src_band = src_ds.GetRasterBand(i+1)
if src_band.GetNoDataValue() is not None:
dst_band = dst_ds.GetRasterBand(i+1)
dst_band.SetNoDataValue(src_band.GetNoDataValue())
dst_band.Fill(src_band.GetNoDataValue())
# reproject the image
dst_ds.SetProjection(dst_sr.ExportToWkt())
dst_ds.SetGeoTransform(tmp_ds.GetGeoTransform())
gdal.ReprojectImage(src_ds, dst_ds,
src_sr.ExportToWkt(),
dst_sr.ExportToWkt(),
gdal.GRA_Bilinear)
tmp_ds = None
# copy the metadata
copy_metadata(src_ds, dst_ds)
return dst_ds
except:
cleanup_temp(dst_ds)
raise
def __call__(self, src_ds):
logger.info("Applying ColorIndexOptimization")
try:
dst_ds = create_temp(src_ds.RasterXSize,
src_ds.RasterYSize,
1,
gdal.GDT_Byte,
temp_root=self.temporary_directory)
if not self.palette_file:
# create a color table as a median of the given dataset
ct = gdal.ColorTable()
gdal.ComputeMedianCutPCT(src_ds.GetRasterBand(1),
src_ds.GetRasterBand(2),
src_ds.GetRasterBand(3), 256, ct)
else:
# copy the color table from the given palette file
pct_ds = gdal.Open(self.palette_file)
pct_ct = pct_ds.GetRasterBand(1).GetRasterColorTable()
if not pct_ct:
raise ValueError("The palette file '%s' does not have a "
"Color Table." % self.palette_file)
ct = pct_ct.Clone()
pct_ds = None
dst_ds.GetRasterBand(1).SetRasterColorTable(ct)
gdal.DitherRGB2PCT(src_ds.GetRasterBand(1),
src_ds.GetRasterBand(2),
src_ds.GetRasterBand(3),
dst_ds.GetRasterBand(1), ct)
copy_projection(src_ds, dst_ds)
copy_metadata(src_ds, dst_ds)
return dst_ds
except:
cleanup_temp(dst_ds)
raise
def __call__(self, src_ds):
logger.info("Applying ColorIndexOptimization")
try:
dst_ds = create_temp(src_ds.RasterXSize, src_ds.RasterYSize,
1, gdal.GDT_Byte,
temp_root=self.temporary_directory)
if not self.palette_file:
# create a color table as a median of the given dataset
ct = gdal.ColorTable()
gdal.ComputeMedianCutPCT(src_ds.GetRasterBand(1),
src_ds.GetRasterBand(2),
src_ds.GetRasterBand(3),
256, ct)
else:
# copy the color table from the given palette file
pct_ds = gdal.Open(self.palette_file)
pct_ct = pct_ds.GetRasterBand(1).GetRasterColorTable()
if not pct_ct:
raise ValueError("The palette file '%s' does not have a "
"Color Table." % self.palette_file)
ct = pct_ct.Clone()
pct_ds = None
dst_ds.GetRasterBand(1).SetRasterColorTable(ct)
gdal.DitherRGB2PCT(src_ds.GetRasterBand(1),
src_ds.GetRasterBand(2),
src_ds.GetRasterBand(3),
dst_ds.GetRasterBand(1), ct)
copy_projection(src_ds, dst_ds)
copy_metadata(src_ds, dst_ds)
return dst_ds
except:
cleanup_temp(dst_ds)
raise
def __call__(self, src_ds):
logger.info("Applying BandSelectionOptimization")
try:
dst_ds = create_temp(src_ds.RasterXSize,
src_ds.RasterYSize,
len(self.bands),
self.datatype,
temp_root=self.temporary_directory)
dst_range = get_limits(self.datatype)
multiple, multiple_written = 0, False
for dst_index, (src_index, dmin, dmax) in enumerate(self.bands, 1):
# check if next band is equal
if dst_index < len(self.bands) and \
(src_index, dmin, dmax) == self.bands[dst_index]:
multiple += 1
continue
# check that src band is available
if src_index > src_ds.RasterCount:
continue
# initialize with zeros if band is 0
if src_index == 0:
src_band = src_ds.GetRasterBand(1)
data = numpy.zeros((src_band.YSize, src_band.XSize),
dtype=gdal_array.codes[self.datatype])
src_min, src_max = (0, 0)
# use src_ds band otherwise
else:
src_band = src_ds.GetRasterBand(src_index)
src_min, src_max = src_band.ComputeRasterMinMax()
# get min/max values or calculate from band
if dmin is None:
dmin = get_limits(src_band.DataType)[0]
elif dmin == "min":
dmin = src_min
if dmax is None:
dmax = get_limits(src_band.DataType)[1]
elif dmax == "max":
dmax = src_max
src_range = (float(dmin), float(dmax))
block_x_size, block_y_size = src_band.GetBlockSize()
num_x = int(math.ceil(float(src_band.XSize) / block_x_size))
num_y = int(math.ceil(float(src_band.YSize) / block_y_size))
dst_band = dst_ds.GetRasterBand(dst_index)
if src_band.GetNoDataValue() is not None:
dst_band.SetNoDataValue(src_band.GetNoDataValue())
for block_x, block_y in product(range(num_x), range(num_y)):
offset_x = block_x * block_x_size
offset_y = block_y * block_y_size
size_x = min(src_band.XSize - offset_x, block_x_size)
size_y = min(src_band.YSize - offset_y, block_y_size)
data = src_band.ReadAsArray(offset_x, offset_y, size_x,
size_y)
# perform clipping and scaling
data = ((dst_range[1] - dst_range[0]) *
((numpy.clip(data, dmin, dmax) - src_range[0]) /
(src_range[1] - src_range[0])))
# set new datatype
data = data.astype(gdal_array.codes[self.datatype])
# write result
dst_band.WriteArray(data, offset_x, offset_y)
# write equal bands at once
if multiple > 0:
for i in range(multiple):
dst_band_multiple = dst_ds.GetRasterBand(
dst_index - 1 - i)
dst_band_multiple.WriteArray(
data, offset_x, offset_y)
multiple_written = True
if multiple_written:
multiple = 0
multiple_written = False
copy_projection(src_ds, dst_ds)
copy_metadata(src_ds, dst_ds)
return dst_ds
except:
cleanup_temp(dst_ds)
raise
def __call__(self, src_ds):
logger.info("Applying BandSelectionOptimization")
try:
dst_ds = create_temp(src_ds.RasterXSize, src_ds.RasterYSize,
len(self.bands), self.datatype,
temp_root=self.temporary_directory)
dst_range = get_limits(self.datatype)
multiple, multiple_written = 0, False
for dst_index, (src_index, dmin, dmax) in enumerate(self.bands, 1):
# check if next band is equal
if dst_index < len(self.bands) and \
(src_index, dmin, dmax) == self.bands[dst_index]:
multiple += 1
continue
# check that src band is available
if src_index > src_ds.RasterCount:
continue
# initialize with zeros if band is 0
if src_index == 0:
src_band = src_ds.GetRasterBand(1)
data = numpy.zeros(
(src_band.YSize, src_band.XSize),
dtype=gdal_array.codes[self.datatype]
)
src_min, src_max = (0, 0)
# use src_ds band otherwise
else:
src_band = src_ds.GetRasterBand(src_index)
src_min, src_max = src_band.ComputeRasterMinMax()
# get min/max values or calculate from band
if dmin is None:
dmin = get_limits(src_band.DataType)[0]
elif dmin == "min":
dmin = src_min
if dmax is None:
dmax = get_limits(src_band.DataType)[1]
elif dmax == "max":
dmax = src_max
src_range = (float(dmin), float(dmax))
block_x_size, block_y_size = src_band.GetBlockSize()
num_x = int(math.ceil(float(src_band.XSize) / block_x_size))
num_y = int(math.ceil(float(src_band.YSize) / block_y_size))
dst_band = dst_ds.GetRasterBand(dst_index)
if src_band.GetNoDataValue() is not None:
dst_band.SetNoDataValue(src_band.GetNoDataValue())
for block_x, block_y in product(range(num_x), range(num_y)):
offset_x = block_x * block_x_size
offset_y = block_y * block_y_size
size_x = min(src_band.XSize - offset_x, block_x_size)
size_y = min(src_band.YSize - offset_y, block_y_size)
data = src_band.ReadAsArray(
offset_x, offset_y, size_x, size_y
)
# perform clipping and scaling
data = ((dst_range[1] - dst_range[0]) *
((numpy.clip(data, dmin, dmax) - src_range[0]) /
(src_range[1] - src_range[0])))
# set new datatype
data = data.astype(gdal_array.codes[self.datatype])
# write result
dst_band.WriteArray(data, offset_x, offset_y)
# write equal bands at once
if multiple > 0:
for i in range(multiple):
dst_band_multiple = dst_ds.GetRasterBand(
dst_index-1-i
)
dst_band_multiple.WriteArray(
data, offset_x, offset_y
)
multiple_written = True
if multiple_written:
multiple = 0
multiple_written = False
copy_projection(src_ds, dst_ds)
copy_metadata(src_ds, dst_ds)
return dst_ds
except:
cleanup_temp(dst_ds)
raise
