Real-Time Video Quality Assessment Without Reference Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video quality assessment methods rely on full-reference models that require access to an original source, making them impractical for end-user devices and inefficient for live video streams with captions, advertisements, and other text or graphics, and are computationally intensive.
Innovation Solution
A machine learning algorithm using compression parameters with auto-tagged continuous variables to assess video quality in real-time, employing transfer learning and distributed machine learning pipelines to extract statistical features from live video streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-reference objective mathematical models are used for video quality assessment, then measurement precision is improved, but device complexity and computational resources increase
Solution Approach 1:
The patent extracts only the essential features needed for quality assessment from the full video signal. Instead of comparing entire video sequences frame-by-frame against a reference, the system extracts key statistical features and compression parameters that capture the most important quality indicators, thereby reducing computational complexity while maintaining assessment accuracy.
Solution Approach 2:
The video quality assessment process is segmented into independent analysis components. The system divides the assessment into separate feature extraction modules (compression parameters, statistical features) that can be processed independently and then combined, reducing the overall computational burden compared to holistic full-reference methods.
2Measurement precision
If full-reference methods are used for video quality assessment, then measurement precision is improved, but ease of operation deteriorates due to source dependency
Solution Approach 1:
The patent implements a self-service quality assessment mechanism that does not require external reference video input. The system uses the video signal itself and its compression parameters to perform quality assessment, making the process autonomous and applicable in scenarios where reference video is unavailable, such as end-user devices and live streaming.
3Measurement precision
If full-reference objective mathematical models are used for video quality assessment, then measurement precision is improved, but productivity decreases due to computational intensity
Solution Approach 1:
The patent applies partial action by assessing quality at selectively chosen moments rather than continuously analyzing every frame. The system evaluates video quality at key points in the stream using compressed representations and statistical features, achieving sufficient quality measurement without the excessive computational effort of frame-by-frame full-reference analysis.
Data Source
AI summary
A method for video quality assessment implemented by a computing device, where the method includes selecting a first set of frames from an input video source, determining features of the first set of frames for input into a first machine learning model, applying the first machine learning model to the features of the first set of frames to obtain a video quality score and a confidence score, and triggering a retraining of the first machine learning model in response to the confidence score being determined to be out of bounds.