Neural Network Video Quality Evaluation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video quality evaluation methods struggle to accurately quantify video quality in the presence of complex distortions introduced by modern video processing techniques and synthetic data generation, limiting their effectiveness in optimizing compression algorithms and rendering techniques.

Innovation Solution

The use of trained neural networks, specifically 3D convolutional neural networks, to evaluate video quality by processing target and reference videos and generating quality metrics and error maps that reflect human perceptual quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If subjective human evaluations are used to assess video quality, then measurement precision is improved, but productivity deteriorates due to time-consuming and costly manual assessment

Engineering Contradiction:

Improvevideo quality assessment accuracyVSAvoidassessment speed and scalability

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent creates a computational model (neural network) that copies and simulates human visual perception mechanisms. The model learns to predict human quality ratings by training on datasets containing both video samples and corresponding subjective human evaluation scores, thereby automating the assessment process while maintaining accuracy comparable to human experts

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical human evaluation process with an automated computational system. Instead of relying on human observers to watch and rate videos, the system uses trained neural networks to process video data and generate quality metrics automatically, eliminating the time and resource constraints of manual assessment

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If traditional objective video quality metrics are used, then productivity is improved through automated assessment, but measurement precision deteriorates when dealing with complex distortions from modern video processing techniques

Engineering Contradiction:

Improveautomated assessment capabilityVSAvoidvideo quality quantification accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent fundamentally changes the parameters and features used for quality assessment. Instead of traditional metrics that focus on compression artifacts and transmission errors, the system employs deep neural network features that capture complex spatiotemporal patterns, including those introduced by modern processing techniques like super-resolution, frame interpolation, and neural rendering

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent combines multiple sources of information into a composite quality assessment model. The system integrates features from different neural network layers, combines multiple quality metrics, and fuses information from both reference and distorted videos to create a comprehensive quality evaluation that handles diverse distortion types

Inventive Principle:

Principle #40Composite materials

3Extent of automation

If automated objective metrics are implemented, then the extent of automation is improved, but measurement precision deteriorates for complex distortions introduced by modern video processing and synthetic data generation

Engineering Contradiction:

Improvevideo quality evaluation automationVSAvoidquality metric accuracy

Core Design Contradiction:

Extent of automationVSMeasurement precision

Solution Approach 1:

The patent performs preliminary training and calibration of the neural network model using extensive datasets with ground truth human evaluations. This preliminary action establishes the model's ability to accurately predict quality before deployment, ensuring that automation maintains high measurement precision across various video processing scenarios

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms where quality assessment results are continuously refined. The model can be retrained and adjusted based on new data and performance metrics, allowing the automated system to improve its measurement precision over time and adapt to emerging video processing techniques

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250200735A1Video quality evaluation based on trained neural networks

Publication Date: 2025.06.19 INTEL CORP

AI summary

Systems, apparatus, articles of manufacture, and methods to evaluate video quality based on trained neural networks are disclosed. An example apparatus disclosed herein obtains, using a trained neural network, target features corresponding to a target video, the target features based on one or more layers of the trained neural network. The example apparatus also obtains, using the trained neural network, reference features corresponding to a reference video, the reference features based on the one or more layers of the trained neural network, the reference video associated with the target video. The example apparatus further outputs a quality metric for the target video based on the target features, the reference features, and a set of weights. In some examples, the apparatus optionally outputs an error map for the target video.