Rate and Distortion: A toolkit for video codec comparison
=========================================================
.. intro-section-start
R&D is a Python library that faciliates multimedia codec comparison. It provides functions to perform rate-distortion analysis for a given codec with sparse samples in the rate-distortion space.
.. intro-section-end
.. feature-section-start
========
Features
========
* Efficiency: R&D provides the implementation of state-of-the-art rate-distortion function models, which enables efficient codec comparison with only sparse encoding samples.
* Accuracy: R&D is demonstrated to be more accuracy than the traditional BD-rate measure.
* Flexibility: R&D can be applied to analyze the encoder performance at different resolution, frame rate, and bit depth, with a unified interface.
* Robustness: R&D curve/surface fitting models are mathematically stable.
* Generalizability: R&D is compatible with most of the modern video quality assessment models such as PSNR, VMAF, and SSIMPlus.
* Extrapolation Capability: R&D can deliver reasonable codec performance analysis in the region where no encoding samples are provided.
.. feature-section-end
.. installation-section-start
============
Installation
============
`R&D <https://github.com/UWIVC/randd>`_ can be installed using ``pip`` or from source.
Install from source:
.. code-block:: sh
$ git clone git@github.com:UWIVC/randd.git
$ python setup.py install
Install from pip:
.. code-block:: sh
$ pip install randd
.. installation-section-end
.. usage-example-start
============
Sample Usage
============
1D Example
----------
The following example illustrates how to use R&D to compare two video codecs with distortion measure being PSNR:
.. code-block:: python
import randd as rd
import numpy as np
r1 = np.array([100., 400., 1500., 6000.])
d1 = np.array([23.83, 25.89, 31.28, 38.22])
r2 = np.array([100., 400., 1500., 6000.])
d2 = np.array([16.25, 29.75, 33.52, 39.12])
analyzer = rd.Analyzer(d_measure='psnr')
quality_gain, bitrate_saving, summary = analyzer(r1, d1, r2, d2, codec1='h264', codec2='vp9')
.. image:: docs/source/_static/example1.png
2D Example
----------
The following example illustrates how to use R&D to compare two video codecs with distortion measure being VMAF. We can also specify the bitrate region [100, 3000] to compute the quality gain. Note that R&D can perform reasonable extrapolation to the region where no RD samples are given.
.. code-block:: python
import randd as rd
import numpy as np
# 1st dim: bitrate, 2nd dim: diagonal size of spatial resolution
r1 = np.array([[2400., 400.], [1700., 865.], [ 300., 1469.], [2100., 2203.]])
d1 = np.array([28.39, 71.87, 26.98, 89.35])
r2 = np.array([[2400., 400.], [1700., 865.], [ 300., 1469.], [2100., 2203.]])
d2 = np.array([28.86, 75.03, 54.35, 96.49])
analyzer = rd.Analyzer(d_measure='vmaf', ndim=2, r_roi=[100, 3000])
quality_gain, bitrate_saving, summary = analyzer(r1, d1, r2, d2, codec1='h264', codec2='vp9')
.. image:: docs/source/_static/example2.png
Backward Compatibility
----------------------
R&D also provides implementation of traditional codec comparison tools such as BD-rate:
.. code-block:: python
import randd as rd
import numpy as np
from randd.model import LogCubic
r1 = np.array([100., 400., 1500., 6000.])
d1 = np.array([23.83, 25.89, 31.28, 38.22])
r2 = np.array([100., 400., 1500., 6000.])
d2 = np.array([16.25, 29.75, 33.52, 39.12])
analyzer = rd.Analyzer(d_measure='psnr', model1=LogCubic, model2=LogCubic)
quality_gain, bitrate_saving, summary = analyzer(r1, d1, r2, d2, codec1='h264', codec2='vp9')
.. image:: docs/source/_static/example3.png
.. usage-example-end
=============
Documentation
=============
The full documentation is available at `<https://randd.readthedocs.io/>`_.
.. citation-section-start
========
Citation
========
We are making R&D publicly available. If you use R&D in your project, we kindly ask you to cite the following paper:
.. code-block:: tex
@article{duanmu2020egrd,
title={Characterizing Generalized Rate-Distortion Performance of Video Coding: An Eigen Analysis Approach},
author={Duanmu, Zhengfang and Liu, Wentao and Li, Zhuoran and Ma, Kede and Wang, Zhou},
journal={IEEE Transactions on Image Processing},
volume={29},
number={},
pages={6180-6193},
year={2020}
}
.. citation-section-end
.. contact-section-start
========
Contacts
========
**Wentao Liu** - `@w238liu <https://ece.uwaterloo.ca/~w238liu>`_ - ``w238liu@uwaterloo.ca``
**Zhengfang Duanmu** - `@zduanmu <https://ece.uwaterloo.ca/~zduanmu>`_ - ``zduanmu@uwaterloo.ca``
.. contact-section-end
Raw data
{
"_id": null,
"home_page": "https://github.com/UWIVC/randd",
"name": "randd",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.10, <4",
"maintainer_email": "",
"keywords": "rate-distortion performance,video encoding",
"author": "Wentao Liu, Zhengfang Duanmu",
"author_email": "liu.wen.tao90@gmail.com, alexduanmu@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/96/5a/a3a326bc909db31b3a48d2565b56db2aa6947f40432e9ac666ae5bf0708a/randd-0.2.2.tar.gz",
"platform": null,
"description": "Rate and Distortion: A toolkit for video codec comparison\n=========================================================\n\n.. intro-section-start\n\nR&D is a Python library that faciliates multimedia codec comparison. It provides functions to perform rate-distortion analysis for a given codec with sparse samples in the rate-distortion space.\n\n.. intro-section-end\n\n.. feature-section-start\n\n========\nFeatures\n========\n\n* Efficiency: R&D provides the implementation of state-of-the-art rate-distortion function models, which enables efficient codec comparison with only sparse encoding samples.\n* Accuracy: R&D is demonstrated to be more accuracy than the traditional BD-rate measure.\n* Flexibility: R&D can be applied to analyze the encoder performance at different resolution, frame rate, and bit depth, with a unified interface.\n* Robustness: R&D curve/surface fitting models are mathematically stable.\n* Generalizability: R&D is compatible with most of the modern video quality assessment models such as PSNR, VMAF, and SSIMPlus.\n* Extrapolation Capability: R&D can deliver reasonable codec performance analysis in the region where no encoding samples are provided.\n\n.. feature-section-end\n\n.. installation-section-start\n\n============\nInstallation\n============\n\n`R&D <https://github.com/UWIVC/randd>`_ can be installed using ``pip`` or from source.\n\nInstall from source:\n\n.. code-block:: sh\n\n $ git clone git@github.com:UWIVC/randd.git\n $ python setup.py install\n\nInstall from pip:\n\n.. code-block:: sh\n\n $ pip install randd\n\n.. installation-section-end\n\n.. usage-example-start\n\n============\nSample Usage\n============\n\n1D Example\n----------\nThe following example illustrates how to use R&D to compare two video codecs with distortion measure being PSNR:\n\n.. code-block:: python\n\n import randd as rd\n import numpy as np\n\n r1 = np.array([100., 400., 1500., 6000.])\n d1 = np.array([23.83, 25.89, 31.28, 38.22])\n r2 = np.array([100., 400., 1500., 6000.])\n d2 = np.array([16.25, 29.75, 33.52, 39.12])\n analyzer = rd.Analyzer(d_measure='psnr')\n quality_gain, bitrate_saving, summary = analyzer(r1, d1, r2, d2, codec1='h264', codec2='vp9')\n\n.. image:: docs/source/_static/example1.png\n\n\n2D Example\n----------\nThe following example illustrates how to use R&D to compare two video codecs with distortion measure being VMAF. We can also specify the bitrate region [100, 3000] to compute the quality gain. Note that R&D can perform reasonable extrapolation to the region where no RD samples are given.\n\n.. code-block:: python\n\n import randd as rd\n import numpy as np\n\n # 1st dim: bitrate, 2nd dim: diagonal size of spatial resolution\n r1 = np.array([[2400., 400.], [1700., 865.], [ 300., 1469.], [2100., 2203.]])\n d1 = np.array([28.39, 71.87, 26.98, 89.35])\n r2 = np.array([[2400., 400.], [1700., 865.], [ 300., 1469.], [2100., 2203.]])\n d2 = np.array([28.86, 75.03, 54.35, 96.49])\n analyzer = rd.Analyzer(d_measure='vmaf', ndim=2, r_roi=[100, 3000])\n quality_gain, bitrate_saving, summary = analyzer(r1, d1, r2, d2, codec1='h264', codec2='vp9')\n\n.. image:: docs/source/_static/example2.png\n\n\nBackward Compatibility\n----------------------\nR&D also provides implementation of traditional codec comparison tools such as BD-rate:\n\n.. code-block:: python\n\n import randd as rd\n import numpy as np\n from randd.model import LogCubic\n\n r1 = np.array([100., 400., 1500., 6000.])\n d1 = np.array([23.83, 25.89, 31.28, 38.22])\n r2 = np.array([100., 400., 1500., 6000.])\n d2 = np.array([16.25, 29.75, 33.52, 39.12])\n analyzer = rd.Analyzer(d_measure='psnr', model1=LogCubic, model2=LogCubic)\n quality_gain, bitrate_saving, summary = analyzer(r1, d1, r2, d2, codec1='h264', codec2='vp9')\n\n.. image:: docs/source/_static/example3.png\n\n.. usage-example-end\n\n=============\nDocumentation\n=============\n\nThe full documentation is available at `<https://randd.readthedocs.io/>`_.\n\n\n.. citation-section-start\n\n========\nCitation\n========\nWe are making R&D publicly available. If you use R&D in your project, we kindly ask you to cite the following paper:\n\n.. code-block:: tex\n\n @article{duanmu2020egrd,\n title={Characterizing Generalized Rate-Distortion Performance of Video Coding: An Eigen Analysis Approach},\n author={Duanmu, Zhengfang and Liu, Wentao and Li, Zhuoran and Ma, Kede and Wang, Zhou},\n journal={IEEE Transactions on Image Processing},\n volume={29},\n number={},\n pages={6180-6193},\n year={2020}\n }\n\n.. citation-section-end\n\n\n.. contact-section-start\n\n========\nContacts\n========\n\n**Wentao Liu** - `@w238liu <https://ece.uwaterloo.ca/~w238liu>`_ - ``w238liu@uwaterloo.ca``\n\n**Zhengfang Duanmu** - `@zduanmu <https://ece.uwaterloo.ca/~zduanmu>`_ - ``zduanmu@uwaterloo.ca``\n\n.. contact-section-end\n",
"bugtrack_url": null,
"license": "MIT License",
"summary": "A Python Library for Video Codec Comparison.",
"version": "0.2.2",
"project_urls": {
"Bug Reports": "https://github.com/UWIVC/randd/issues",
"Homepage": "https://github.com/UWIVC/randd",
"Source": "https://github.com/UWIVC/randd/"
},
"split_keywords": [
"rate-distortion performance",
"video encoding"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "edc403591a6bfd18181bc3ea3794881154741f1e90c5dc72484e4dc7b92d18d5",
"md5": "f3280a0db40ca07d3c177d921dc04c98",
"sha256": "ad4f84432184e8362b47861d73e7de60173e66cc509bc36208024a175f5a0fba"
},
"downloads": -1,
"filename": "randd-0.2.2-py3-none-any.whl",
"has_sig": false,
"md5_digest": "f3280a0db40ca07d3c177d921dc04c98",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.10, <4",
"size": 9966090,
"upload_time": "2024-01-31T18:40:57",
"upload_time_iso_8601": "2024-01-31T18:40:57.194652Z",
"url": "https://files.pythonhosted.org/packages/ed/c4/03591a6bfd18181bc3ea3794881154741f1e90c5dc72484e4dc7b92d18d5/randd-0.2.2-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "965aa3a326bc909db31b3a48d2565b56db2aa6947f40432e9ac666ae5bf0708a",
"md5": "c353fd9938a13b8ed1d0d4ff1ecc01d7",
"sha256": "ba023b63be8dcbebad1bf16704069ec417c4e360d73b3cefca5964f709ba98d9"
},
"downloads": -1,
"filename": "randd-0.2.2.tar.gz",
"has_sig": false,
"md5_digest": "c353fd9938a13b8ed1d0d4ff1ecc01d7",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.10, <4",
"size": 11307826,
"upload_time": "2024-01-31T18:41:08",
"upload_time_iso_8601": "2024-01-31T18:41:08.049983Z",
"url": "https://files.pythonhosted.org/packages/96/5a/a3a326bc909db31b3a48d2565b56db2aa6947f40432e9ac666ae5bf0708a/randd-0.2.2.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-01-31 18:41:08",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "UWIVC",
"github_project": "randd",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [],
"tox": true,
"lcname": "randd"
}
JSON
