Video Quality Indicators

Here you can download the indicators. You can download them for free, use it for research, but please refer to its original source (corresponding paper + link to this site).

Nawała, Jakub; Janowski, Lucjan; Leszczuk, Mikołaj

Modeling of quality of experience in no-reference model Journal Article

In: Journal of Telecommunications and Information Technology, 2017.

BibTeX

If you refer to individual indicators, please cite specific papers from the table below.

Python Package

The project is available as Python package that performs all the calculations on the video input file.

pip install agh-vqis

Single Executable:

The project is also available as one executable file that performs all the calculations on the .yuv input file or as separated indicators implemented in MATLAB.

Original Image

Bash/Batch Script:

Apart from the single executable file that calculates indicators using the RAW videos, you can download here a BASH/BATCH script that utilizes the ffmpeg tools in order to process whole folders of videos saved with normal extensions like: .avi, .mkv, etc.

The only prerequisite is to have the ffmpeg tools installed.

Bash script (Linux/Mac OS)

Batch script (Windows)

Python Script:

If you need to process the large amount of YUV videos and create the large spread sheet comprising all the results you can download here the python script written by Fredrik Pihl.

In order to use it you need to install the “xlsxwriter” Python module and rename all the video files according to the following convention: NAME_WIDTHxHEIGHT_FPS_….yuv. The last prerequisite is to have the “mitsuLinuxMultithread” binary in the same directory as this script.

See the source code

MATLAB Scripts: (DEPRECATED)

What is more, we are providing you with each separate indicator implemented as the MATLAB script.

Although it is present here, we do not recommend to use this version of the indicators.

Download the MATLAB scripts

List of Implemented Indicators:

Commercial Black
Blockiness
Block Loss
Blur
Contrast
Exposure
Flickering
Freezing
Interlacing
Letter-boxing
Noise
Pillar-boxing
Slicing
Spatial Activity
Temporal Activity
Colourfulness
Blur Amount
User-Generated Content

Indicators	Range of results (min/max)	Monotonicity relative to distortion	The value for image without distortion	Comment
Commercial Black	min. = 0 (blackout did not occur) max. = 1 (blackout occurred)	Greater value → greater distortion (in this case distortion occurs)	0	Indicator treats every unitary frame as blackout (regardless of color).
Blockiness	min. = 0 max. = 3570 (values goes to infinity therefore the results were limited)	Greater value → less visible distortion	From 0.9 to 1.01
Block Loss	min. = 0 max. ≈ 100 – 200	Greater value → more visible distortion	From 0 to 5	Indicator returns -1 if the image has a width or height less than 256 pixels.
Blur	min. = 0 max. ≈ 70	Greater value → more visible distortion	From 0 to 5
Contrast	min. = 0 max. ≈ 120	Greater value → higher contrast	From 45 to 55
Exposure	min. = 0 max. = 255	Greater value → greater exposure time	From 115 to 125
Flickering	min. = 0 max. = 8	Greater value → more visible distortion	For time window with a length of 8 frames typical value is around 0.125	Indicator operates on observation window = number of consecutive frames taken into account when calculating flicker blocks indicator. Result is returned only at the end of the observation window, during the window, the indicator returns -1.
Freezing	min. = 0 max. = 1	Greater value → greater distortion (in this case distortion occurs)	0	Indicator is coupled with the results of Temporal Activity indicator.
Interlacing	min. = 0 max. = 1	Greater value → greater distortion	0
Letter-boxing	min. = 0 max. = 1	Greater value → greater distortion	0	Value 1 means that the entire frame is smooth (blackout).
Noise	min. = 0 max. = 30	Greater value → greater distortion	From 0 to 3.5
Pillar-boxing	min. = 0 max. = 1	Greater value → greater distortion	0	Value 1 means that the entire frame is smooth (blackout).
Slicing	min. ≈ 0 max. = &infin		≈ 0	Indicator does not work correctly.
Spatial Activity	min. = 0 max. ≈ 270	Greater value → greater Spatial Activity	From 0 to 60
Temporal Activity	min. = 0 max. = 255 (for fullHD)	Greater value → greater Temporal Activity	From 0 to 20	Range of results depends on size of the test image.
Colourfulness	min. = 0 max. ≈ 170	Greater value → more colourful	None	Indicating amount of colourfulness.
Blur Amount	min. = 0 max. = 1	Greater value → less blurred	1.0	Computing the cumulative probability of blur detection.
User-Generated Content	min. = 0 max. = 1	1 → UGC 0 → professional (non-UGC)	1	Indicating whether content was professionally generated or not.
Lip Sync				Coming soon!

List of implemented indicators

Video Indicators

Not all video errors are detected as the indicator list does not include all possible indicators. Moreover, after obtaining the values of the indicators, it is recommended to analyze their time series. It is also worth noting that there is currently no support for resolution changes (although this support is under preparation).