Blosc2 grok

A plugin of the excellent grok library for Blosc2. grok is a JPEG2000 codec, and with this plugin, you can use it as yet another codec in applications using Blosc2. See an example of use at: https://github.com/Blosc/blosc2_grok/blob/main/examples/params.py

Installation

For using blosc2_grok you will first have to install its wheel:

pip install blosc2-grok -U

Usage

import blosc2
import numpy as np
import blosc2_grok
from PIL import Image

# Set the params for the grok codec
kwargs = {}
kwargs['cod_format'] = blosc2_grok.GrkFileFmt.GRK_FMT_JP2
kwargs['quality_mode'] = "dB"
kwargs['quality_layers'] = np.array([5], dtype=np.float64)
blosc2_grok.set_params_defaults(**kwargs)

# Define the compression and decompression parameters for Blosc2.
# Disable the filters and do not split blocks (these won't work with grok).
cparams = {
    'codec': blosc2.Codec.GROK,
    'filters': [],
    'splitmode': blosc2.SplitMode.NEVER_SPLIT,
}

# Read the image
im = Image.open("examples/kodim23.png")
# Convert the image to a numpy array
np_array = np.asarray(im)

# Transform the numpy array to a blosc2 array. This is where compression happens, and
# the HTJ2K codec is called.
bl_array = blosc2.asarray(
    np_array,
    chunks=np_array.shape,
    blocks=np_array.shape,
    cparams=cparams,
    urlpath="examples/kodim23.b2nd",
    mode="w",
)

# Print information about the array, see the compression ratio (cratio)
print(bl_array.info)

Parameters for compression

The following parameters are available for compression for grok, with their defaults. Most of them are named after the ones in the Pillow library and have the same meaning. The ones that are not in Pillow are marked with a * and you can get more information about them in the grok documentation, or by following the provided links. For those marked with a **, you can get more information in the grok.h header.

'tile_size': (0, 0),
'tile_offset': (0, 0),
'quality_mode': None,
'quality_layers': np.zeros(0, dtype=np.float64),
'progression': "LRCP",
'num_resolutions': 6,
'codeblock_size': (64, 64),
'irreversible': False,
'precinct_size': (0, 0),
'offset': (0, 0),
'mct': 0,
* 'numgbits': 2,  # Equivalent to -N, -guard_bits
* 'roi_compno': -1,  # Together with 'roi_shift' it is equivalent to -R, -ROI
* 'roi_shift': 0,
* 'decod_format': GrkFileFmt.GRK_FMT_UNK,
* 'cod_format': GrkFileFmt.GRK_FMT_UNK,
* 'rsiz': GrkProfile.GRK_PROFILE_NONE,  # Equivalent to -Z, -rsiz
* 'framerate': 0,
* 'apply_icc_': False,  # Equivalent to -f, -apply_icc
* 'rateControlAlgorithm': GrkRateControl.BISECT,
* 'num_threads': 0,
* 'deviceId': 0,  # Equivalent to -G, -device_id
* 'duration': 0,  # Equivalent to -J, -duration
* 'repeats': 1,  # Equivalent to -e, -repetitions
* 'mode': GrkMode.DEFAULT,  # Equivalent to -M, -mode
* 'verbose': False,  # Equivalent to -v, -verbose
** 'enableTilePartGeneration': False,  # See header of grok.h above
** 'max_cs_size': 0,  # See header of grok.h above
** 'max_comp_size': 0,  # See header of grok.h above

*Note: * when using the blosc2_grok plugin from C, the structure used for setting the parameters uses the grok parameters names. You can see an example in https://github.com/Blosc/leaps-examples/blob/main/c-compression/compress-tomo.c#L110 .

codec_meta as rates quality mode

As a simpler way to activate the rates quality mode, if you set the codec_meta from the cparams to an integer different from 0, the rates quality mode will be activated with a rate value equal to codec_meta / 10. If cod_format is not specified, the default will be used. The codec_meta has priority to the rates param set with the blosc2_grok.set_params_defaults(). Please note that only rates < 25.6 are supported with this notation.

import blosc2


cparams = {
    'codec': blosc2.Codec.GROK,
    'codec_meta': 5 * 10,  # cratio will be 5
    'filters': [],
    'splitmode': blosc2.SplitMode.NEVER_SPLIT,
}

Notes

When using blosc2_grok, there are some restrictions that you have to keep in mind.

• The minimum supported image size is around 256 bytes, so an image with less size will fail to be compressed.
• The maximum datatype precision is of 16 bits.
• Although floats from 16 or fewer bits of precision seem to work, we recommend using integer data when possible.

More examples

See the examples directory for more examples.

Thanks

Thanks to Marta Iborra, from the Blosc Development Team, for doing most of the job in making this plugin possible, and J. David Ibáñez and Francesc Alted for the initial contributions. Also, thanks to Aaron Boxer, the original author of the grok library, for his help in ironing out issues for making this interaction possible.

That's all folks!

The Blosc Development Team