SSIM-Based Rate-Distortion Optimization Model for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as HEVC and VVC, face challenges in optimizing structure similarity index (SSIM) for efficient video compression, particularly in balancing bitrate and distortion, especially with the introduction of new technologies like VVC which aims to achieve higher compression efficiency.
Innovation Solution
The development of a rate-distortion optimization (RDO) model trained using SSIM or MS-SSIM indices, which generates training data from video sequences to process and optimize video coding, allowing for better bitrate and distortion management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional video coding standards (HEVC, VVC) are used with traditional rate-distortion optimization, then coding efficiency is improved, but structure similarity index (SSIM) optimization is insufficient
Solution Approach 1:
The patent changes the optimization parameter from traditional rate-distortion optimization to SSIM-based rate-distortion optimization. By training a new RDO model using SSIM as the distortion metric instead of traditional metrics like MSE, the system achieves better structure similarity preservation while maintaining coding efficiency. This parameter change allows the encoder to prioritize perceptual quality over mathematical accuracy.
2Measurement precision
If SSIM-based rate-distortion optimization is implemented, then structure similarity index is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training the rate-distortion optimization model using SSIM metrics before actual video encoding. The training phase computes SSIM values and establishes the relationship between bitrate and SSIM for different quantization parameters. This pre-computed knowledge is then stored and reused during encoding, avoiding the need to compute SSIM in real-time during the actual video processing, thus reducing computational complexity while maintaining SSIM optimization.
3Speed
If traditional rate-distortion optimization is used, then computational speed is maintained, but compression performance and subjective quality assessment are insufficient
Solution Approach 1:
The patent substitutes the traditional mechanical rate-distortion optimization mechanism with a learned model-based approach. Instead of using fixed formulas and heuristics for RDO, the system employs a neural network model trained on SSIM metrics to predict the optimal quantization parameters. This substitution allows the system to maintain computational speed through efficient model inference while achieving superior compression performance and subjective quality assessment aligned with human visual perception.
Data Source
AI summary
The present disclosure provides a computer-implemented method for encoding video. The method includes: generating training data based on one or more video sequences, the training data including a structure similarity index comprising at least one of structure similarity index (SSIM) or multi-scale-structural similarity index (MS-SSIM); training a rate-distortion optimization (RDO) model using the training data; processing the one or more video sequences using the rate-distortion optimization model.