Research Project
Constant Subjective Quality Database
This project developed a novel measurement platform and methodology for generating video sequences of constant subjective quality, supporting perceptual video assessment, adaptive codecs, and machine learning-based video quality research.
Key Results
Research Overview
Video databases with subjective assessments form an essential foundation for learning algorithms, objective video quality assessment, and the development of adaptive video compression technologies. This project introduces a novel device and methodology for measuring coded video quality through acceptable minimum perception thresholds.
The proposed system generates video sequences of continuous subjective quality using a specialised installation that dynamically changes the visual quality of typical media content during the examination process. This approach enables a much larger volume of accurately labelled perceptual data than traditional binary estimates.
The resulting database contains approximately 4.1 million video quality perceptual thresholds. These data support the development of effective adaptive codecs, hybrid video models, machine learning-based quality metrics, and future perceptual video processing systems.
Research Contributions
Novel Measurement Platform
- Dynamic visual quality control during the subjective examination process.
- Generation of constant subjective quality sequences using controlled perceptual thresholds.
- Specialised experimental installation for measuring coded video quality.
- Continuous perceptual assessment instead of limited binary quality estimates.
Scientific Outcomes
- 4.1 million perceptual thresholds for video quality research.
- Large-scale labelled dataset for objective metric development.
- Foundation for adaptive codecs and hybrid video compression models.
- Training data for machine learning and AI-driven video quality systems.
Experimental System
Publications, Code and Data
EMCTECH 2021
Mozhaeva A, Potashnikov A, Vlasuyk I and Streeter L (2021). Constant Subjective Quality Database: The Research and Device of Generating Video Sequences of Constant Quality. International Conference on Engineering Management of Communication and Technology, Vienna, Austria.
IVCNZ 2023
Mozhaeva A, Mazin V, Cree MJ and Streeter L (2023). NRspttemVQA: Real-Time Video Quality Assessment Based on the User’s Visual Perception. 38th International Conference on Image and Vision Computing New Zealand.
Code and Data
Source code and data related to the Constant Subjective Quality Database are available in the project repository.
Dataset Access
Access to the complete annotated dataset is available upon request for research and collaboration purposes.
Applications
Contact
Primary contact: amozhaeva@eit.ac.nz.