Block-Based Video Quality Assessment Using Sensitivity Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for evaluating video quality require excessive time to model and approximate a human's visual system, and there is a need for a more efficient method to transmit a minimum amount of information while maintaining perceived quality.
Innovation Solution
A video quality assessment network is used to split images into blocks, determine sensitivity and quality assessment information, and combine these to evaluate final image quality, with deep neural networks for downscaling and upscaling to enhance encoding and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a human visual system model is used to evaluate video quality, then measurement precision is improved, but evaluation time increases excessively
Solution Approach 1:
The patent creates a simplified copy of the human visual system through a trained neural network model that replicates HVS-based quality assessment behavior. Instead of implementing the complex original HVS model, a lightweight neural network is trained to copy its quality assessment functionality, achieving similar measurement precision with significantly reduced computational time and complexity.
Solution Approach 2:
The patent replaces the mechanical/computational complexity of the human visual system model with a data-driven neural network approach. By substituting the traditional HVS modeling methodology with a trained deep learning model, the system achieves comparable quality assessment accuracy while dramatically reducing processing time and computational resources required.
2Productivity
If deep learning techniques are applied to video quality assessment, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent segments the video quality assessment task into distinct processing stages: block division, feature extraction through convolutional layers, quality score prediction, and temporal aggregation. This segmentation allows each component to be optimized independently, managing overall system complexity while maintaining high productivity through specialized sub-functions.
Solution Approach 2:
The patent implements dynamic adaptability through trainable neural network parameters that are optimized during training and then fixed for efficient inference. The system transitions from a complex training phase with adjustable parameters to a simplified deployment phase where the trained model operates with fixed weights, reducing runtime complexity while maintaining high assessment speed.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Proposed are a method and apparatus for evaluating the quality of an image, the method including obtaining blocks each having a predetermined size by splitting a target image for evaluating a quality and a reference image that is to be compared with the target image, determining sensitivity information and quality assessment information of each of the blocks by inputting the blocks to a video quality assessment network, and determining a final image quality assessment score of the target image by combining the pieces of quality assessment information of the blocks with each other, based on the pieces of sensitivity information of the blocks.