Perceived Media Quality Prediction Using Adversarial Annotations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for compressing media content, such as images and videos, often result in degraded quality due to bandwidth constraints and the need for compression, making it difficult to predict and optimize perceived quality effectively.
Innovation Solution
A network-accessible media optimization service uses a machine learning model that combines scores from multiple perceptual quality algorithms to predict perceived image quality and an iterative evolutionary approach to optimize compression algorithms, selecting hyper-parameters that balance file size and quality, thereby improving user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If compression techniques are used to reduce network bandwidth consumption and latency, then network efficiency and speed are improved, but the perceived quality of media objects deteriorates
Solution Approach 1:
The system employs perceptual quality algorithms that provide feedback about the perceived quality of compressed media objects. This feedback loop allows the system to adjust compression parameters to maintain acceptable quality levels while optimizing network efficiency, directly resolving the contradiction between compression efficiency and quality preservation.
Solution Approach 2:
The system dynamically changes compression parameters based on multiple perceptual quality algorithm scores. By adjusting parameters such as compression ratio, quality factor, and encoding settings, the system optimizes the balance between file size reduction and perceived quality maintenance, resolving the technical contradiction between network efficiency and quality.
2Measurement precision
If multiple perceptual quality algorithms are used to predict quality, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The system merges multiple perceptual quality algorithm scores into a unified quality prediction. By combining the strengths of different algorithms (e.g., SSIM for structural similarity, PSNR for signal-to-noise ratio, and other perceptual metrics), the system achieves higher prediction accuracy than any single algorithm could provide alone, while managing computational complexity through efficient integration.
3Quantity of substance
If compression is applied to reduce file size, then storage requirements are reduced, but the extent of perceived degradation becomes harder to predict
Solution Approach 1:
The system performs preliminary quality assessment using multiple perceptual algorithms before final compression is applied. By predicting perceived degradation in advance and adjusting compression parameters accordingly, the system ensures that quality remains within acceptable thresholds while achieving the desired file size reduction, thus maintaining prediction precision despite compression.
Data Source
AI summary
Respective labels indicative of compression-related quality degradation for a set of media object tuples which meet a divergence criterion are obtained; each tuple comprises a reference media object and a pair of corresponding compressed media object versions. Pairs of training records for a machine learning model are generated using the labeled media object tuples and multiple perceptual quality algorithms, with each training record comprising respective perceived quality degradation scores generated by each of the multiple algorithms for a given compressed media object of a tuple. A machine learning model is trained, using the record pairs, to predict quality degradation scores for compressed media objects.