JP2Grok -- JPEG2000 driver based on Grok library

Driver short name

JP2Grok

Build dependencies

Grok >= 20.2

Added in version 3.13.

This driver is an implementation of a JPEG2000 reader/writer based on the Grok library.

The driver uses the VSI Virtual File API, so it can read JPEG2000 compressed NITF files. It also supports native S3 access through Grok's built-in cloud I/O.

XMP metadata can be extracted from JPEG2000 files, and will be stored as XML raw content in the xml:XMP metadata domain.

The driver supports writing georeferencing information as GeoJP2 and GMLJP2 boxes.

The driver supports creating files with transparency, arbitrary band count, and adding/reading metadata. Update of georeferencing or metadata of existing file is also supported. Optional intellectual property metadata can be read/written in the xml:IPR box.

Driver capabilities

Supports CreateCopy()

This driver supports the GDALDriver::CreateCopy() operation

Supports Georeferencing

This driver supports georeferencing

Supports VirtualIO

This driver supports virtual I/O operations (/vsimem/, etc.)

Georeferencing

Georeferencing information can come from different sources : internal (GeoJP2 or GMLJP2 boxes), worldfile .j2w/.wld sidecar files, or PAM (Persistent Auxiliary metadata) .aux.xml sidecar files. By default, information is fetched in following order (first listed is the highest priority): PAM, GeoJP2, GMLJP2, WORLDFILE.

The allowed sources and their priority order can be changed with the GDAL_GEOREF_SOURCES configuration option (or GEOREF_SOURCES open option) whose value is a comma-separated list of the following keywords : PAM, GEOJP2, GMLJP2, INTERNAL (shortcut for GEOJP2,GMLJP2), WORLDFILE, NONE. Earlier mentioned sources take priority over later ones. A non mentioned source will be ignored.

For example setting it to "WORLDFILE,PAM,INTERNAL" will make a geotransformation matrix from a potential worldfile priority over PAM or internal JP2 boxes. Setting it to "PAM,WORLDFILE,GEOJP2" will use the mentioned sources and ignore GMLJP2 boxes.

Thread support

The Grok library uses multi-threaded decompression internally via its own thread pool. Multi-threaded encoding is also supported. The number of threads is determined at runtime from GDALGetNumThreads.

When a JPEG2000 file contains multiple tiles, GDAL will additionally dispatch tile processing across threads for parallel CopyTileData operations during decoding.

Performance: AdviseRead

The driver decompresses asynchronously, scheduling tile decode in the background while the caller is still consuming earlier rows. To do this efficiently it needs to know the read window up front so it can restrict the codestream decode to the targeted precincts/tiles and overlap T2 parse with T1 decode.

For best performance, callers that know which window they are about to read should call GDALDataset::AdviseRead (or the C API GDALDatasetAdviseRead) with that window before issuing the first RasterIO / ReadBlock. Without the hint the driver falls back to synchronous, per-swath decompression, which can be much slower, especially when decompressing over the network.

Open Options

Open options can be specified in command-line tools using the syntax -oo <NAME>=<VALUE> or by providing the appropriate arguments to GDALOpenEx() (C) or gdal.OpenEx (Python). The following open options are available:

1BIT_ALPHA_PROMOTION=[YES/NO]: Defaults to YES. Whether a 1-bit alpha channel should be promoted to 8-bit.
GEOREF_SOURCES=value: Define which georeferencing sources are allowed and their priority order. See Georeferencing paragraph.
USE_TILE_AS_BLOCK=[YES/NO]: Defaults to NO. Whether to always use the JPEG-2000 block size as the GDAL block size. Setting this option can be useful when doing whole image decompression and the image is single-tiled. Note however that the tile size must not exceed 2 GB since that's the limit supported by GDAL.

Creation Options

Creation options can be specified in command-line tools using the syntax -co <NAME>=<VALUE> or by providing the appropriate arguments to GDALCreate() (C) or Driver.Create (Python). The following creation options are available:

CODEC=[JP2/J2K]: JP2 will add JP2 boxes around the codestream data. The value is determined automatically from the file extension. If it is neither JP2 nor J2K, J2K codec is used.
GMLJP2=[YES/NO]: Defaults to YES. Indicates whether a GML box conforming to the OGC GML in JPEG2000 specification should be included in the file. Unless GMLJP2V2_DEF is used, the version of the GMLJP2 box will be version 1.
GMLJP2V2_DEF=[YES/<filename>/<json>]: Indicates whether a GML box conforming to the OGC GML in JPEG2000, version 2.0.1 specification should be included in the file. filename must point to a file with a JSON content that defines how the GMLJP2 v2 box should be built. See the JP2OpenJPEG -- JPEG2000 driver based on OpenJPEG library driver documentation for the syntax of the JSON configuration file. It is also possible to directly pass the JSON content inlined as a string. If filename is just set to YES, a minimal instance will be built.
GeoJP2=[YES/NO]: Defaults to YES. Indicates whether a UUID/GeoTIFF box conforming to the GeoJP2 (GeoTIFF in JPEG2000) specification should be included in the file.
QUALITY=<float_value,float_value,...>: Percentage between 0 and 100. A value of 50 means the file will be half-size in comparison to uncompressed data, 33 means 1/3, etc.. Defaults to 25 (unless the dataset is made of a single band with color table, in which case the default quality is 100). It is possible to specify several quality values (comma separated) to ask for several quality layers. Quality values should be increasing.
REVERSIBLE=[YES/NO]: YES means use of reversible 5x3 integer-only filter, NO use of the irreversible DWT 9-7. Defaults to NO (unless the dataset is made of a single band with color table, in which case reversible filter is used).
RESOLUTIONS=<integer>: Number of resolution levels. Default value is selected such the smallest overview of a tile is no bigger than 128x128.
BLOCKXSIZE=<integer>: Defaults to 1024. Tile width.
BLOCKYSIZE=<integer>: Defaults to 1024. Tile height.
PROGRESSION=[LRCP/RLCP/RPCL/PCRL/CPRL]: Defaults to LRCP. Progression order.
SOP=[YES/NO]: Defaults to NO. YES means generate SOP (Start Of Packet) marker segments.
EPH=[YES/NO]: Defaults to NO. YES means generate EPH (End of Packet Header) marker segments.
YCBCR420=[YES/NO]: Defaults to NO. YES if RGB must be resampled to YCbCr 4:2:0.
YCC=[YES/NO]: Defaults to YES. YES if RGB must be transformed to YCC color space ("MCT transform", i.e. internal transform, without visual degradation).
NBITS=<integer>: Bits (precision) for sub-byte files (1-7), sub-uint16 (9-15), sub-uint32 (17-31).
1BIT_ALPHA=[YES/NO]: Whether to encode the alpha channel as a 1-bit channel (when there's an alpha channel). Defaults to NO, unless INSPIRE_TG=YES. Enabling this option might cause compatibility problems with some readers.
ALPHA=[YES/NO]: Defaults to NO. Whether to force encoding last channel as alpha channel. Only useful if the color interpretation of that channel is not already Alpha.
PROFILE=[AUTO/UNRESTRICTED/PROFILE_1]: Defaults to AUTO. Determine which codestream profile to use. UNRESTRICTED corresponds to the "Unrestricted JPEG 2000 Part 1 codestream" (RSIZ=0). PROFILE_1 corresponds to the "JPEG 2000 Part 1 Profile 1 codestream" (RSIZ=2), which add constraints on tile dimensions and number of resolutions. In AUTO mode, the driver will determine if the BLOCKXSIZE, BLOCKYSIZE, RESOLUTIONS, CODEBLOCK_WIDTH and CODEBLOCK_HEIGHT values are compatible with PROFILE_1 and advertise it in the relevant case. Note that the default values of those options are compatible with PROFILE_1. Otherwise UNRESTRICTED is advertised.
INSPIRE_TG=[YES/NO]: Defaults to NO. Whether to use JPEG2000 features that comply with Inspire Orthoimagery Technical Guidelines. When set to YES, implies PROFILE=PROFILE_1, 1BIT_ALPHA=YES, GEOBOXES_AFTER_JP2C=YES. The CODEC, BLOCKXSIZE, BLOCKYSIZE, RESOLUTIONS, NBITS, PROFILE, CODEBLOCK_WIDTH and CODEBLOCK_HEIGHT options will be checked against the requirements and recommendations of the Technical Guidelines.
JPX=[YES/NO]: Defaults to YES. Whether to advertise JPX features, and add a Reader requirement box, when a GMLJP2 box is written. This option should not be used unless compatibility problems with a reader occur.
GEOBOXES_AFTER_JP2C=[YES/NO]: Whether to place GeoJP2/GMLJP2 boxes after the code-stream. Defaults to NO, unless INSPIRE_TG=YES. This option should not be used unless compatibility problems with a reader occur.
PRECINCTS=[<{prec_w,prec_h},{prec_w,prec_h}/...>]: A list of {precincts width,precincts height} tuples to specify precincts size. Each value should be a multiple of 2. The maximum number of tuples used will be the number of resolutions. The first tuple corresponds to the higher resolution level, and the following ones to the lower resolution levels. If less tuples are specified, the last one is used by dividing its values by 2 for each extra lower resolution level. The default value used is {512,512},{256,512},{128,512},{64,512},{32,512},{16,512},{8,512},{4,512},{2,512}. An empty string may be used to disable precincts ( i.e. the default {32767,32767},{32767,32767}, ... will then be used).
TILEPARTS=[DISABLED/RESOLUTIONS/LAYERS/COMPONENTS]: Defaults to DISABLED. Whether to generate tile-parts and according to which criterion.
CODEBLOCK_WIDTH=<integer>: Defaults to 64. Codeblock width: power of two value between 4 and 1024. Note that CODEBLOCK_WIDTH * CODEBLOCK_HEIGHT must not be greater than 4096. For PROFILE_1 compatibility, CODEBLOCK_WIDTH must not be greater than 64.
CODEBLOCK_HEIGHT=<integer>: Defaults to 64. Codeblock height: power of two value between 4 and 1024. Note that CODEBLOCK_WIDTH * CODEBLOCK_HEIGHT must not be greater than 4096. For PROFILE_1 compatibility, CODEBLOCK_HEIGHT must not be greater than 64.
CODEBLOCK_STYLE=value: Style of the code-block coding passes. The following 6 independent settings can be combined together (values should be comma separated):
- BYPASS (1): enable selective arithmetic coding bypass (can substantially improve coding/decoding speed, at the expense of larger file size)
- RESET (2): reset context probabilities on coding pass boundaries
- TERMALL (4): enable termination on each coding pass
- VSC (8): enable vertically causal context
- PREDICTABLE (16): enable predictable termination (helps for error detection)
- SEGSYM (32): enable segmentation symbols (helps for error detection)
Instead of specifying them by text, it is also possible to give the corresponding numeric value of the global codeblock style, by adding the selected options (for example "BYPASS,TERMALL" is equivalent to "5"=1+4)

By default, none of them are enabled. Enabling them will generally increase codestream size, but improve either coding/decoding speed or resilience/error detection.
PLT=[YES/NO]: Defaults to NO. Whether to write a PLT (Packet Length) marker segment in tile-part headers.
TLM=[YES/NO]: Defaults to NO. Whether to write a TLM (Tile-part Length) marker segment in main header.
WRITE_METADATA=[YES/NO]: Defaults to NO. Whether metadata should be written, in a dedicated JP2 'xml ' box. The content of the 'xml ' box will be like:
```
<GDALMultiDomainMetadata>
  <Metadata>
    <MDI key="foo">bar</MDI>
  </Metadata>
  <Metadata domain='aux_domain'>
    <MDI key="foo">bar</MDI>
  </Metadata>
  <Metadata domain='a_xml_domain' format='xml'>
    <arbitrary_xml_content>
    </arbitrary_xml_content>
  </Metadata>
</GDALMultiDomainMetadata>
```
If there are metadata domain whose name starts with "xml:BOX_", they will be written each as separate JP2 'xml ' box.

If there is a metadata domain whose name is "xml:XMP", its content will be written as a JP2 'uuid' XMP box.

If there is a metadata domain whose name is "xml:IPR", its content will be written as a JP2 'jp2i' box.
MAIN_MD_DOMAIN_ONLY=[YES/NO]: Defaults to NO. (Only if WRITE_METADATA=YES) Whether only metadata from the main domain should be written.
USE_SRC_CODESTREAM=[YES/NO]: (EXPERIMENTAL!) When source dataset is JPEG2000, whether to reuse the codestream of the source dataset unmodified. Defaults to NO. Note that enabling that feature might result in inconsistent content of the JP2 boxes w.r.t. to the content of the source codestream. Most other creation options will be ignored in that mode. Can be useful in some use cases when adding/correcting georeferencing, metadata, ...
TRANSCODE=[YES/NO]: Defaults to NO. When the source dataset is JPEG 2000, transcode the codestream without a full decompression/recompression round-trip. The underlying codestream data is copied verbatim while the following options may be used to modify markers or progression:
- PLT – insert PLT (Packet Length, Tile-part) marker segments
- TLM – insert TLM (Tile-part Length) marker segments
- SOP – insert SOP (Start Of Packet) marker segments
- EPH – insert EPH (End of Packet Header) marker segments
- PROGRESSION – change the progression order (LRCP, RLCP, RPCL, PCRL, CPRL)
JP2 metadata boxes (GeoTIFF UUID, XMP, GMLJP2, IPR, GDAL xml) are regenerated from the current dataset metadata.

Because the codestream coefficients are copied rather than re-encoded, creation options that would affect the pixel data (QUALITY, REVERSIBLE, BLOCKXSIZE, BLOCKYSIZE, RESOLUTIONS, YCBCR420, YCC, NBITS, 1BIT_ALPHA, PRECINCTS, TILEPARTS, CODEBLOCK_WIDTH, CODEBLOCK_HEIGHT, CODEBLOCK_STYLE, USE_SRC_CODESTREAM) are ignored — a warning is emitted if any are supplied.

Example:
```
gdal_translate -of JP2Grok -co TRANSCODE=YES \
    -co PLT=YES -co TLM=YES -co PROGRESSION=RPCL \
    input.jp2 output.jp2
```
COMMENT=<string>: Content of the comment (COM) marker in the codestream.

Lossless compression

Lossless compression can be achieved if ALL the following creation options are defined :

QUALITY=100
REVERSIBLE=YES
YCBCR420=NO (which is the default)

Standalone plugin compilation

While this driver may be built as part of a whole GDAL build, either in libgdal itself, or as a plugin, it is also possible to only build this driver as a plugin, against an already built libgdal.

The version of the GDAL sources used to build the driver must match the version of the libgdal it is built against.

For example, from a "build_grok" directory under the root of the GDAL source tree:

cmake -S ../frmts/grok -DCMAKE_PREFIX_PATH=/path/to/GDAL_installation_prefix -DGROK_ROOT=/path/to/grok_install_prefix
cmake --build .

Note that such a plugin, when used against a libgdal not aware of it, will be systematically loaded at GDAL driver initialization time, and will not benefit from deferred plugin loading capabilities <rfc-96>. For that, libgdal itself must be built with the CMake variable GDAL_REGISTER_DRIVER_JP2GROK_FOR_LATER_PLUGIN=ON set.