agh-vqis

agh-vqis

A Python package computing a set of image quality indicators (IQIs) for a given input video.

Following IQIs are included in the package:

• A set of 15 Video Quality Indicators (VQIs) developed by the team from AGH. See the following website for more information: https://qoe.agh.edu.pl/indicators/.
• Our Python reimplementation of the Colourfulness IQI. See this paper for more information regarding this IQI.
• Blur Amount IQI. See package's Python Package Index web page for more information.
• UGC IQI (User-generated content).

Authors: Jakub Nawała <jakub.nawala@agh.edu.pl>, Filip Korus <fkorus@student.agh.edu.pl>

Requirements

• ffmpeg - version >= 4.4.2
• Python - version >= 3.9

Installation

pip install agh_vqis

Usage

1. Single multimedia file:

   from agh_vqis import process_single_mm_file, VQIs
   from pathlib import Path
   
   if __name__ == '__main__':
       process_single_mm_file(Path('/path/to/single/video.mp4'))
   

2. Folder with multimedia files:

   from agh_vqis import process_folder_w_mm_files, VQIs
   from pathlib import Path
   
   if __name__ == '__main__':
       process_folder_w_mm_files(Path('/path/to/multimedia/folder/'))
   

3. Options parameter - in either process_single_mm_file and process_folder_w_mm_files function options could be provided as an optional argument. It is being passed to function as a dictionary like below.

    process_single_mm_file(Path('/path/to/single/video.mp4'), options={
         VQIs.contrast: False,  # disable contrast indicator
         VQIs.colourfulness: False,  # disable colourfulness indicator
    })
   

4. How to disable/enable indicators to count? Every indicator is enabled by default except blur_amount due to long computing time. To disable one of following indicators (blockiness, SA, letterbox, pillarbox, blockloss, blur, TA, blackout, freezing, exposure, contrast, interlace, noise, slice, flickering, colourfulness, ugc) pass

   VQIs.indicator_name: False
   

   to options dictionary. Whereas to enable blur_amount indicator pass True value.

5. agh-vqis package could be used from command line interface as well. For example:

   python3 -m agh_vqis /path/to/single/movie.mp4 # will run VQIS for single file
   

   or

   python3 -m agh_vqis /path/to/multimedia/folder/ # will run VQIS for folder
   

   Whereas this command will display help:

   $ python3 -m agh_vqis -h
   

6. Supported multimedia files: mp4, y4m, mov, mkv, avi, ts, webm, jpg, jpeg, png, gif, bmp.

7. First row of the output CSV file contains header with image quality indicators (IQIs) names, whereas each row below the header represents single video frame from the input video file. Each column provides a numerical result as returned by a given IQI when applied to the respective frame.

8. Cast chosen indicators for different resolutions (experimental). For example: to cast Blur to 1440p and Blockiness to 2160p you should pass two additional lines in options dictionary like below.

       from agh_vqis import process_single_mm_file, CastVQI, DestResolution
       from pathlib import Path
       
       if __name__ == '__main__':
           process_single_mm_file(Path('/path/to/single/video.mp4'), options={
              CastVQI.blur: DestResolution.p1440,
              CastVQI.blockiness: DestResolution.p2160
           })
   

   Available casts (with percentage of correctness)

   Blockiness:

   source resolution	destination resolution	correctness
   1080p	1440p	99.80%
   1080p	2160p	99.72%
   1440p	2160p	99.68%
   2160p	1440p	99.81%

   Blur:

   source resolution	destination resolution	correctness
   240p	360p	93.91%
   240p	480p	87.78%
   360p	240p	94.28%
   360p	480p	98.08%
   360p	720p	92.51%
   480p	240p	87.68%
   480p	360p	98.05%
   480p	720p	97.35%
   720p	360p	92.69%
   720p	480p	97.31%
   720p	1080p	80.95%
   1080p	1440p	99.12%
   1080p	2160p	93.41%
   1440p	1080p	99.07%
   1440p	2160p	96.24%
   2160p	1080p	93.59%
   2160p	1440p	96.39%

   Exposure(bri):

   source resolution	destination resolution	correctness
   240p	360p	97.75%
   240p	480p	94.89%
   240p	720p	89.71%
   360p	240p	97.77%
   360p	480p	98.68%
   360p	720p	94.83%
   360p	1080p	80.37%
   480p	240p	94.82%
   480p	360p	98.68%
   480p	720p	97.80%
   480p	1080p	82.90%
   480p	1440p	80.40%
   720p	240p	89.32%
   720p	360p	94.87%
   720p	480p	97.85%
   720p	1080p	86.99%
   720p	1440p	84.66%
   720p	2160p	82.34%
   1080p	360p	80.71%
   1080p	480p	83.41%
   1080p	720p	86.88%
   1080p	1440p	98.78%
   1080p	2160p	96.91%
   1440p	480p	80.90%
   1440p	720p	84.85%
   1440p	1080p	98.83%
   1440p	2160p	99.05%
   2160p	720p	82.89%
   2160p	1080p	96.93%
   2160p	1440p	98.98%

   Contrast:

   source resolution	destination resolution	correctness
   240p	360p	99.83%
   240p	480p	99.69%
   240p	720p	99.55%
   240p	1080p	87.94%
   240p	1440p	87.94%
   240p	2160p	87.91%
   360p	240p	99.85%
   360p	480p	99.97%
   360p	720p	99.93%
   360p	1080p	88.36%
   360p	1440p	88.34%
   360p	2160p	88.36%
   480p	240p	99.73%
   480p	360p	99.97%
   480p	720p	99.98%
   480p	1080p	88.61%
   480p	1440p	88.67%
   480p	2160p	88.67%
   720p	240p	99.61%
   720p	360p	99.93%
   720p	480p	99.98%
   720p	1080p	88.45%
   720p	1440p	88.46%
   720p	2160p	88.46%
   1080p	240p	87.12%
   1080p	360p	87.71%
   1080p	480p	87.82%
   1080p	720p	87.91%
   1080p	1440p	99.99%
   1080p	2160p	99.99%
   1440p	240p	87.13%
   1440p	360p	87.72%
   1440p	480p	87.85%
   1440p	720p	87.92%
   1440p	1080p	99.99%
   1440p	2160p	100.0%
   2160p	240p	87.11%
   2160p	360p	87.72%
   2160p	480p	87.85%
   2160p	720p	87.91%
   2160p	1080p	99.99%
   2160p	1440p	100.0%

   Interlace:

   source resolution	destination resolution	correctness
   240p	360p	86.93%
   360p	240p	87.19%
   360p	480p	88.32%
   480p	360p	87.22%
   480p	720p	90.80%
   720p	480p	91.21%
   720p	1080p	82.65%
   1080p	1440p	81.35%

   Noise:

   source resolution	destination resolution	correctness
   240p	360p	88.85%
   360p	240p	88.08%
   360p	480p	86.33%
   480p	360p	85.94%
   480p	720p	88.71%
   720p	480p	88.12%
   1080p	1440p	94.84%
   1080p	2160p	80.34%
   1440p	1080p	92.28%
   1440p	2160p	87.24%
   2160p	1440p	88.54%

License

The agh-vqis Python package is provided via the Evaluation License Agreement.