Bitstream Video Quality Assessment for Encoding Profile Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video streaming services face challenges in determining optimal encoding parameters for diverse content characteristics, especially with codecs like AVC, HEVC, and AV1, which have numerous encoding parameters, affecting video quality and user experience, and conventional subjective quality assessment methods are expensive and impractical for live streaming.
Innovation Solution
A bitstream-based video quality assessment (VQA) model using machine learning to predict quality scores without decoding frames, incorporating features from encoded bitstreams, including metadata and pixel information, to optimize encoding profiles and bitrate settings for various content types and playback devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective quality assessment methods are used to determine video quality, then accuracy of quality evaluation is improved, but cost and practicality for live streaming deteriorate
Solution Approach 1:
The patent creates an objective copy of human visual perception through machine learning models trained on subjective quality data. The VQA model learns to predict human quality scores by analyzing bitstream features, effectively copying human quality assessment capability without requiring actual human subjects. This resolves the contradiction by providing accurate quality evaluation (matching subjective methods) while eliminating the cost and impracticality of live human testing.
2Measurement precision
If full decoding of video frames is performed for quality assessment, then accuracy of quality measurement is improved, but computational complexity and processing time deteriorate
Solution Approach 1:
The patent extracts only the essential quality-related features directly from the bitstream without performing full video decoding. By identifying and extracting relevant encoding parameters, motion vectors, and other bitstream features that correlate with perceived quality, the system achieves accurate quality assessment while avoiding the computational burden of complete frame decoding. This selective extraction resolves the contradiction between measurement accuracy and computational complexity.
3Manufacturing precision
If numerous encoding parameters are adjusted to optimize video quality, then video quality is improved, but difficulty of parameter optimization and system complexity deteriorate
Solution Approach 1:
The patent transforms the complex multi-parameter optimization problem into a simpler prediction task by training machine learning models on encoding parameters and quality scores. Instead of manually optimizing numerous encoding parameters, the system uses the trained VQA model to predict quality outcomes and guide parameter selection. This parameter transformation approach resolves the contradiction by maintaining video quality optimization while significantly reducing system complexity and making the process more manageable.
Data Source
AI summary
Techniques are described for training and use of machine learning models to determine objective video quality scores. Video quality scores predict the quality of video content perceived by viewers. Quality scores have various uses, including the selection of encoding profiles and determination of encoding ladders. A core model and residual model may be used to determine quality scores.