PPD-Aware Perceptual Video Quality Modeling Across Displays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional perceptual quality models fail to accurately predict perceived video quality when viewing parameters differ from those used during training, leading to inaccurate quality/bitrate tradeoffs and reduced overall visual quality in streamed videos.
Innovation Solution
A trained perceptual quality model is developed using machine learning to account for different viewing parameters, incorporating features like VIF indices and a PPD-aware DLM to estimate perceptual quality across varying display resolutions and normalized viewing distances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional perceptual quality model is trained based on a single set of viewing parameters, then the model can accurately predict quality for those specific parameters, but the model cannot accurately predict quality when viewing parameters differ from training parameters
Solution Approach 1:
The patent applies parameter changes by incorporating viewing parameter-dependent features (such as PPD-aware DLM and VIF indices computed at multiple spatial scales) into the quality model. These features dynamically adjust based on the actual viewing parameters (display resolution, viewing distance, PPD values), allowing the model to accurately predict quality across different viewing conditions without requiring separate models for each parameter set.
Solution Approach 2:
The patent achieves universality by creating a single perceptual quality model that can handle multiple viewing parameter configurations. The model uses a unified feature computation framework that adapts to different display resolutions and viewing distances through PPD-aware processing, making one model applicable across diverse viewing scenarios rather than requiring parameter-specific models.
2Quantity of substance
If lossy data compression algorithms are used to reduce encoded video size, then the compression ratio increases, but visual impairments and distortions increase reducing perceived video quality
Solution Approach 1:
The patent implements feedback by using the trained perceptual quality model to evaluate reconstructed video quality and provide guidance to the encoding process. The model predicts perceptual quality scores based on features extracted from the reconstructed video and viewing parameters, allowing the encoding system to adjust compression settings to maintain acceptable quality levels while achieving efficient compression.
Solution Approach 2:
The patent applies parameter changes by using PPD-aware feature computation that adapts to different viewing conditions. The VIF indices and DLM features are computed at spatial scales corresponding to the actual PPD values, allowing the quality assessment to accurately reflect perceived quality across different resolutions and viewing distances, thereby enabling more accurate quality-bitrate tradeoff decisions.
Data Source
AI summary
In various embodiments, a training application generates a trained perceptual quality model that estimates perceived video quality for reconstructed video. The training application computes a pixels-per-degree value based on a normalized viewing distance and a display resolution. The training application computes a set of feature values corresponding to a set of visual quality metrics based on a reconstructed video sequence, a source video sequence, and the pixels-per-degree value. The training application executes a machine learning algorithm on the first set of feature values to generate the trained perceptual quality model. The trained perceptual quality model maps a particular set of feature values corresponding to the set of visual quality metrics to a particular perceptual quality score.