Feature image config

README.rst

@@ -49,7 +49,7 @@ Orthority command line functionality is accessed with the ``oty`` command, and i
 -  ``exif``: Orthorectify images with frame camera model(s) defined by image EXIF / XMP tags.
 -  ``odm``: Orthorectify images in a processed OpenDroneMap dataset that includes a DSM.
 -  ``rpc``: Orthorectify images with RPC camera models defined by image tags / sidecar files or parameter files.
--  ``sharpen``: Pan sharpen an image using the Gram-Schmidt method.
+-  ``sharpen``: Pan-sharpen an image using the Gram-Schmidt method.
 
 Get help on ``oty`` with:
 
@@ -110,7 +110,7 @@ As above, but refine the RPC camera model with GCPs in ``source.tif`` tags:
 
    oty rpc --dem dem.tif --gcp-refine tags source.tif
 
-Pan sharpen the multispectral image ``ms.tif`` with the panchromatic image ``pan.tif``:
+Pan-sharpen the multispectral image ``ms.tif`` with the panchromatic image ``pan.tif``:
 
 .. code-block:: bash
 
@@ -144,7 +144,7 @@ Orthorectify an image using interior and exterior parameter files to generate th
     ortho.process('ortho.tif')
 
 
-Pan sharpen a multispectral image with the matching panchromatic image:
+Pan-sharpen a multispectral image with the matching panchromatic image:
 
 .. below copied from docs/scripts/api_pan_sharp.py
 
@@ -157,7 +157,7 @@ Pan sharpen a multispectral image with the matching panchromatic image:
     pan_file = url_root + 'pan_sharp/pan.tif'  # panchromatic drone image
     ms_file = url_root + 'pan_sharp/ms.tif'  # multispectral (RGB) drone image
 
-    # create PanSharpen object and pan sharpen
+    # create PanSharpen object and pan-sharpen
     pan_sharp = oty.PanSharpen(pan_file, ms_file)
     pan_sharp.process('pan_sharp.tif')
 

docs/api/pan_sharp.rst

@@ -1,4 +1,4 @@
-Pan sharpening
+Pan-sharpening
 ===============
 
 .. automodule:: orthority.pan_sharp

docs/getting_started/api/pan_sharp.rst

@@ -1,7 +1,7 @@
-Pan sharpening
+Pan-sharpening
 ==============
 
-The :class:`~orthority.pan_sharp.PanSharpen` class implements Gram-Schmidt pan sharpening.  Panchromatic and multispectral images are required to instantiate.  The :meth:`~orthority.pan_sharp.PanSharpen.process` method pan sharpens:
+The :class:`~orthority.pan_sharp.PanSharpen` class implements Gram-Schmidt pan-sharpening.  Panchromatic and multispectral images are required to instantiate.  The :meth:`~orthority.pan_sharp.PanSharpen.process` method pan-sharpens:
 
 .. literalinclude:: ../../scripts/api_pan_sharp.py
     :language: python

docs/getting_started/cli/ortho_config.rst

@@ -4,11 +4,14 @@
 Ortho image configuration
 =========================
 
-Ortho images are created as GeoTIFFs.  Image resolution and format (data type, compression, nodata / internal mask and overviews) can be configured.  Configuration options are common to all |oty|_ orthorectification sub-commands and default to sensible values when not supplied.
+Ortho image resolution and format can be configured.  Configuration options are common to all |oty|_ orthorectification sub-commands and default to sensible values when not supplied.
+
+Resolution, data type and compression
+-------------------------------------
 
 Ortho resolution defaults to an estimate of the `ground sampling distance <https://en.wikipedia.org/wiki/Ground_sample_distance>`__.  This can be changed with ``--res``.  The ortho data type defaults to the source image data type, and can be changed with ``--dtype``.
 
-Compression can be configured with ``--compress`` as either ``deflate`` (with any ortho data type) or ``jpeg`` (with the ``uint8`` or ``uint16`` ortho data types).  If ``--compress`` is not specified, compression defaults to ``jpeg`` when the ortho data type is ``uint8``, otherwise it defaults to ``deflate``.  When ``jpeg`` compression is used with the ``uint16`` data type, the ortho is 12 bit ``jpeg`` compressed.
+Compression can be configured with ``--compress`` as either ``deflate`` or ``lzw`` (with any ortho data type), or ``jpeg`` (with the ``uint8`` or ``uint16`` ortho data types).  If ``--compress`` is not specified, compression defaults to ``jpeg`` when the ortho data type is ``uint8``, and to ``deflate`` otherwise.  When ``jpeg`` compression is used with the ``uint16`` data type, the ortho is 12 bit ``jpeg`` compressed.
 
 .. note::
 
@@ -20,11 +23,30 @@ The next example orthorectifies using EXIF / XMP tags, and configures the ortho
 
     oty exif --dem odm/odm_dem/dsm.tif --res 0.2 --dtype uint8 --compress deflate odm/images/100_0005_0140.tif
 
-Valid ortho pixels are masked with either an internal mask band or a nodata value.  By default, an internal mask is used when the ortho image is ``jpeg`` compressed.  This avoids ``jpeg`` artefacts in invalid areas.  When the ortho is ``deflate`` compressed, the default is use to a nodata value based on the data type.  Masking behaviour can be changed with ``--write-mask`` to write an internal mask, or ``--no-write-mask`` to use a nodata value.
+Masking and overviews
+---------------------
+
+Valid ortho pixels are masked with either an internal mask band or a nodata value.  By default, an internal mask is used when the ortho image is ``jpeg`` compressed.  This avoids ``jpeg`` artefacts in invalid areas.  When the ortho is ``deflate`` or ``lzw`` compressed, the default is use to a nodata value based on the data type.  Masking behaviour can be changed with ``--write-mask`` to write an internal mask, or ``--no-write-mask`` to use a nodata value.
 
-Internal overviews are added by default.  This can be changed with ``--no-build-ovw``.  In this example, we create an ortho image with ``deflate`` compression, nodata rather than internal masking, and no internal overviews:
+Internal overviews are added by default.  This can be changed with ``--no-build-ovw``.  In this example, we create an ortho image with ``deflate`` compression, internal masks, and no internal overviews:
 
 .. code-block:: bash
 
-    oty exif --dem odm/odm_dem/dsm.tif --compress deflate --no-write-mask --no-build-ovw odm/images/100_0005_0140.tif
+    oty exif --dem odm/odm_dem/dsm.tif --compress deflate --write-mask --no-build-ovw odm/images/100_0005_0140.tif
+
+Custom creation options and driver
+----------------------------------
+
+For ortho image configurations not possible with the above options, custom creation options can be specified with ``--creation-option``.  The ``--compress`` option is ignored, and no other creation options are set by Orthority when this is supplied.
+
+The ortho can be formatted as a ``gtiff`` (GeoTIFF - the default) or ``cog`` (Cloud Optimised GeoTIFF) with ``--driver``.
+
+This example formats the ortho as a GeoTIFF with internal masks, and specifies custom creation options for tiled YCbCr JPEG compression with a quality of 90:
+
+.. code-block:: bash
+
+    oty exif --dem odm/odm_dem/dsm.tif --write-mask --driver gtiff --creation-option tiled=yes --creation-option compress=jpeg --creation-option photometric=ycbcr --creation-option jpeg_quality=90 odm/images/100_0005_0140.tif
+
+.. note::
 
+    Each driver has its own creation options.  See the GDAL `GeoTIFF <https://gdal.org/en/latest/drivers/raster/gtiff.html#creation-options>`__ and `COG <https://gdal.org/en/latest/drivers/raster/cog.html#creation-options>`__ docs for details.

docs/getting_started/cli/pan_sharpening.rst

@@ -1,7 +1,7 @@
 .. include:: ../../shared.txt
 .. include:: shared.txt
 
-Pan sharpening
+Pan-sharpening
 ==============
 
 .. toctree::

docs/getting_started/cli/sharp_config.rst

@@ -4,24 +4,48 @@
 Output image configuration
 ==========================
 
-Pan sharpened images are created as GeoTIFFs.  Image format (data type, compression, nodata / internal mask and overviews) can be configured.  Configuration options are shared with the |oty|_ orthorectification sub-commands, and default to sensible values when not supplied.
+The pan-sharpened image format can be configured.  Configuration options are shared with the |oty|_ orthorectification sub-commands, and default to sensible values when not supplied.
 
-The image data type defaults to the multispectral image data type, and can be changed with ``--dtype``.  Compression can be configured with ``--compress`` as either ``deflate`` (with any image data type) or ``jpeg`` (with the ``uint8`` or ``uint16`` image data types).  If ``--compress`` is not specified, compression defaults to ``jpeg`` when the image data type is ``uint8``, otherwise it defaults to ``deflate``.  When ``jpeg`` compression is used with the ``uint16`` data type, the image is 12 bit ``jpeg`` compressed.
+Data type and compression
+-------------------------
+
+The image data type defaults to the multispectral image data type, and can be changed with ``--dtype``.  Compression can be configured with ``--compress`` as either ``deflate`` or ``lzw`` (with any image data type), or ``jpeg`` (with the ``uint8`` or ``uint16`` image data types).  If ``--compress`` is not specified, compression defaults to ``jpeg`` when the image data type is ``uint8``, and to ``deflate`` otherwise.  When ``jpeg`` compression is used with the ``uint16`` data type, the image is 12 bit ``jpeg`` compressed.
 
 .. note::
 
     Support for 12 bit JPEG compression is Rasterio_ build / package dependent.
 
-The next example creates a pan sharpened image with the ``int16`` data type, and ``deflate`` compression:
+The next example creates a pan-sharpened image with the ``int16`` data type, and ``deflate`` compression:
 
 .. code-block:: bash
 
     oty sharpen --pan pan_sharp/pan.tif --multispectral pan_sharp/ms.tif --out-file pan_sharp.tif --dtype int16 --compress deflate
 
-Valid image pixels are masked with either an internal mask band or a nodata value.  By default, an internal mask is used when the image image is ``jpeg`` compressed.  This avoids ``jpeg`` artefacts in invalid areas.  When the image is ``deflate`` compressed, the default is use to a nodata value based on the data type.  Masking behaviour can be changed with ``--write-mask`` to write an internal mask, or ``--no-write-mask`` to use a nodata value.
+Masking and overviews
+---------------------
+
+Valid image pixels are masked with either an internal mask band or a nodata value.  By default, an internal mask is used when the image is ``jpeg`` compressed.  This avoids ``jpeg`` artefacts in invalid areas.  When the image is ``deflate`` or ``lzw`` compressed, the default is use to a nodata value based on the data type.  Masking behaviour can be changed with ``--write-mask`` to write an internal mask, or ``--no-write-mask`` to use a nodata value.
 
-Internal overviews are added by default.  This can be changed with ``--no-build-ovw``.  In this example, we create an pan sharpened image with ``deflate`` compression, internal masking rather than nodata, and no internal overviews:
+Internal overviews are added by default.  This can be changed with ``--no-build-ovw``.  In this example, we create an pan-sharpened image with ``deflate`` compression, internal masking rather than nodata, and no internal overviews:
 
 .. code-block:: bash
 
     oty sharpen --pan pan_sharp/pan.tif --multispectral pan_sharp/ms.tif --out-file pan_sharp.tif --compress deflate --write-mask --no-build-ovw
+
+Custom creation options and driver
+----------------------------------
+
+For pan-sharpened image configurations not possible with the above options, custom creation options can be specified with ``--creation-option``.  The ``--compress`` option is ignored, and no other creation options are set by Orthority when this is supplied.
+
+The image can be formatted as a ``gtiff`` (GeoTIFF - the default) or ``cog`` (Cloud Optimised GeoTIFF) with ``--driver``.
+
+This example formats the pan-sharpened image as a COG with internal masks, and specifies custom creation options for JPEG compression with a quality of 90:
+
+.. code-block:: bash
+
+    oty sharpen --pan pan_sharp/pan.tif --multispectral pan_sharp/ms.tif --out-file pan_sharp.tif --write-mask --driver cog --creation-option compress=jpeg --creation-option quality=90
+
+.. note::
+
+    Each driver has its own creation options.  See the GDAL `GeoTIFF <https://gdal.org/en/latest/drivers/raster/gtiff.html#creation-options>`__ and `COG <https://gdal.org/en/latest/drivers/raster/cog.html#creation-options>`__ docs for details.
+

docs/getting_started/cli/sharpen.rst

@@ -3,7 +3,7 @@
 ``oty sharpen``
 ===============
 
-|oty sharpen|_ implements the Gram-Schmidt pan sharpening method.  This example pan sharpens a multispectral (RGB) drone image with its matching panchromatic image:
+|oty sharpen|_ implements the Gram-Schmidt pan-sharpening method.  This example pan-sharpens a multispectral (RGB) drone image with its matching panchromatic image:
 
 .. code-block:: bash
 

docs/scripts/api_frame.py

@@ -1,4 +1,3 @@
-# TODO: change URL to main branch
 # code for getting started->api->camera models->frame cameras
 # [create camera]
 import orthority as oty

docs/scripts/api_pan_sharp.py

@@ -5,6 +5,6 @@
 pan_file = url_root + 'pan_sharp/pan.tif'  # panchromatic drone image
 ms_file = url_root + 'pan_sharp/ms.tif'  # multispectral (RGB) drone image
 
-# create PanSharpen object and pan sharpen
+# create PanSharpen object and pan-sharpen
 pan_sharp = oty.PanSharpen(pan_file, ms_file)
 pan_sharp.process('pan_sharp.tif')

orthority/__init__.py

@@ -25,7 +25,7 @@
 import logging
 import pathlib
 
-from orthority.enums import Compress, Interp, RpcRefine
+from orthority.enums import Compress, Interp, RpcRefine, Driver
 from orthority.factory import FrameCameras, RpcCameras
 from orthority.ortho import Ortho
 from orthority.pan_sharp import PanSharpen

orthority/camera.py

@@ -217,7 +217,7 @@ def rect_boundary(im_size: np.ndarray, num_pts: int) -> np.ndarray:
             perim = 2 * br.sum()
             cnr_ji = np.array([[0, 0], [br[0], 0], br, [0, br[1]], [0, 0]])
             dist = np.sum(np.abs(np.diff(cnr_ji, axis=0)), axis=1)
-            return np.row_stack(
+            return np.vstack(
                 [
                     np.linspace(
                         cnr_ji[i],
@@ -754,7 +754,7 @@ def _get_rectangles(
             n = 9
             scale_j, scale_i = np.meshgrid(range(0, n), range(0, n))
             scale_j, scale_i = scale_j.ravel(), scale_i.ravel()
-            ji = np.row_stack([scale_j * w / (n - 1), scale_i * h / (n - 1)])
+            ji = np.vstack([scale_j * w / (n - 1), scale_i * h / (n - 1)])
             xy = self._pixel_to_camera(ji)[:2]
             outer = xy.min(axis=1), xy.max(axis=1) - xy.min(axis=1)
             inner_ul = np.array((xy[0][scale_j == 0].max(), xy[1][scale_i == 0].max()))
@@ -791,7 +791,7 @@ def _camera_to_pixel(self, xyz_: np.ndarray) -> np.ndarray:
 
     def _pixel_to_camera(self, ji: np.ndarray) -> np.ndarray:
         """Transform 2D pixel to homogenous 3D camera coordinates."""
-        ji_ = np.row_stack([ji.astype('float64', copy=False), np.ones((1, ji.shape[1]))])
+        ji_ = np.vstack([ji.astype('float64', copy=False), np.ones((1, ji.shape[1]))])
         xyz_ = self._K_undistort_inv.dot(ji_)
         return xyz_
 
@@ -1237,7 +1237,7 @@ def _get_undistort_maps(self) -> tuple[np.ndarray, np.ndarray]:
         # equivalent to the above, but using Camera methods (works out slower):
         # j = np.arange(0, self.im_size[0], dtype='int32')
         # i = np.zeros(self.im_size[0], dtype='int32')
-        # ji = np.row_stack((j, i))
+        # ji = np.vstack((j, i))
         # undistort_maps = (
         #     np.zeros(self.im_size[::-1], dtype='float32'),
         #     np.zeros(self.im_size[::-1], dtype='float32'),
@@ -1258,7 +1258,7 @@ def _camera_to_pixel(self, xyz_: np.ndarray) -> np.ndarray:
     def _pixel_to_camera(self, ji: np.ndarray) -> np.ndarray:
         ji_cv = ji.T.astype('float64', copy=False)
         xyz_ = cv2.undistortPoints(ji_cv, self._K, self._dist_param)
-        xyz_ = np.row_stack([xyz_[:, 0, :].T, np.ones((1, ji.shape[1]))])
+        xyz_ = np.vstack([xyz_[:, 0, :].T, np.ones((1, ji.shape[1]))])
         return xyz_
 
 
@@ -1480,7 +1480,7 @@ def _camera_to_pixel(self, xyz_: np.ndarray) -> np.ndarray:
     def _pixel_to_camera(self, ji: np.ndarray) -> np.ndarray:
         ji_cv = ji.T[None, :].astype('float64', copy=False)
         xyz_ = cv2.fisheye.undistortPoints(ji_cv, self._K, self._dist_param, None, None)
-        xyz_ = np.row_stack([xyz_[0].T, np.ones((1, ji.shape[1]))])
+        xyz_ = np.vstack([xyz_[0].T, np.ones((1, ji.shape[1]))])
         return xyz_