Content-Dependent Video Quality Model for Streaming Services
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video quality estimation models fail to accurately consider the content-dependent quality impact of digital video signals, especially in encrypted streams, and do not provide a fine-grained measurement of spatio-temporal complexity, which affects perceived quality during packet-loss and non-packet-loss scenarios.
Innovation Solution
A method that extracts content-dependent parameters from Group Of Picture (GOP)/scene-complexity parameters, using packet-header information, to estimate the perceived quality of digital video signals, allowing for a fine-grained consideration of content impact and applicability to both encrypted and non-encrypted streams, including cases with and without packet-loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If parameter-based video quality estimation models are used, then automation and efficiency are improved, but content-dependent quality impact is not accurately considered
Solution Approach 1:
The patent introduces content-complexity parameters (spatial complexity Cs and temporal complexity Ct) as additional input parameters to the existing parameter-based quality estimation model. These parameters are extracted from the video bitstream and used to modulate the quality estimation, thereby incorporating content-dependent effects while maintaining the automated nature of the assessment.
Solution Approach 2:
The patent creates a composite quality estimation approach by combining multiple parameter types: network parameters (packet loss rate, bitrate), coding parameters (quantization parameter, motion vector magnitude), and newly introduced content-complexity parameters (spatial and temporal complexity). This composite parameter set enables both automation and content-dependent accuracy.
2Device complexity
If coarse threshold-based complexity measurement is used, then device complexity is reduced, but measurement precision of content impact is insufficient
Solution Approach 1:
The patent replaces coarse threshold-based complexity classification with continuous parameter measurements. Instead of categorizing frames as simply 'complex' or 'simple' based on thresholds, the model computes continuous spatial complexity (Cs) and temporal complexity (Ct) values that provide fine-grained differentiation of content characteristics.
Solution Approach 2:
The patent transitions from one-dimensional threshold classification to two-dimensional continuous parameter space by introducing both spatial complexity (Cs) and temporal complexity (Ct) as separate continuous dimensions. This enables more nuanced quality estimation by considering multiple aspects of content complexity simultaneously.
3Adaptability or versatility
If relative complexity measurement within video sequence is used, then adaptation to content variations is improved, but comparison across different contents becomes impossible
Solution Approach 1:
The patent transforms relative complexity measurements into absolute measurements by normalizing the complexity parameters (Cs and Ct) against reference values or standard scales. This allows the model to adapt to different content types while producing comparable quality estimates across diverse video contents through standardized parameter ranges.
Data Source
AI summary
A method for estimating the perception quality of a digital video signal includes: (1a) extracting information of the video bit stream, which is captured prior to decoding; (1b) getting estimation(s) for one or more impairment factors IF using, for each of the estimations, an impact function adapted for the respective impairment factor; and (1c) estimating the perceived quality of the digital video signal using the estimation(s) obtained in step (1b).