Video Quality Estimation Using Encrypted Frame Size Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating user-perceived quality of video streams fail to accurately account for the underlying video content, especially when the stream is encrypted, leading to inaccurate network adjustments and degraded performance.
Innovation Solution
A method that dynamically estimates user-perceived quality by analyzing the chronological sequence of absolute or relative sizes of encoded video frames, using reference patterns or statistical measures to identify characteristics associated with content-dependent variations, even if the video stream payload is encrypted.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If packet-layer parameters are used to estimate quality, then the estimation can be performed dynamically and objectively, but the accuracy is insufficient because video content characteristics are not captured
Solution Approach 1:
The patent uses frame size as an intermediary parameter that can be extracted from encrypted packet-layer data without decryption. Frame size serves as a mediator that preserves video content characteristics (motion, detail, artifacts) while remaining accessible in encrypted streams, thereby resolving the contradiction between dynamic estimation capability and accuracy.
2Measurement precision
If decoded video frames are evaluated directly, then accurate quality measurement is achieved, but the approach cannot be used when the video stream payload is encrypted
Solution Approach 1:
The patent extracts frame size information from the encrypted video stream at the packet layer without requiring decryption of the payload. This extraction approach obtains content-dependent characteristics (motion, detail, artifacts) while maintaining compatibility with encrypted streams, thus resolving the contradiction between measurement accuracy and adaptability to encrypted content.
3Measurement precision
If subjective user surveys are conducted, then accurate Mean Opinion Score is obtained, but large human resources and time are required preventing real-time application
Solution Approach 1:
The patent replaces the mechanical system of human subjective evaluation with an automated objective analysis system that processes frame size sequences. This substitution maintains accuracy by capturing content-dependent quality characteristics while eliminating time consumption and human resource requirements, enabling real-time quality monitoring.
Data Source
AI summary
One or more embodiments herein dynamically and objectively estimate user-perceived quality of an encoded video stream in dependence on the underlying video content of that stream. Estimation processing according to these embodiments includes generating a chronological sequence of the absolute or relative sizes of encoded video frames in the stream. Processing further entails analyzing this sequence to identify a plurality of reference characteristics that are defined in a reference model. These reference characteristics are more specifically defined in the reference model as parameters that characterize or are otherwise associated with content-dependent variations in user-perceived quality. For example, the reference characteristics in some embodiments comprise reference patterns or reference statistical measures that have been determined based on analyzing frame size sequences generated for reference video streams with known qualities. Regardless, processing finally includes estimating the user-perceived quality of the stream based on the identified reference characteristics, according to the reference model.