ML Prediction Coding Models for More Accurate Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video coding technologies face inefficiencies due to inaccurate ad-hoc prediction models, limiting the compression efficiency of digital data transmission and storage.
Innovation Solution
Implementing machine learning prediction coding models, such as trained artificial neural networks, to optimize prediction coding by iteratively partitioning training data and satisfying defined convergence criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ad-hoc prediction models are used for video coding, then device complexity is reduced and ease of operation is improved, but manufacturing precision and measurement precision deteriorate due to inaccurate prediction
Solution Approach 1:
The prediction coding models are trained offline using training data partitions before deployment. The training process pre-computes optimal model parameters and convergence criteria, so that during actual video coding operation, the pre-trained models can be directly applied without real-time training complexity. This separates the complex training phase from the simple inference phase.
Solution Approach 2:
The system automatically partitions training data and trains prediction coding models without requiring manual intervention for each coding scenario. The machine learning framework self-adjusts model parameters through iterative training processes, enabling the system to adapt to different video content characteristics autonomously while maintaining reasonable computational complexity.
2Measurement precision
If machine learning prediction coding models are implemented, then prediction accuracy is improved, but device complexity and computational requirements increase
Solution Approach 1:
The training data is partitioned into multiple subsets, and different prediction coding models are trained on different partitions. This segmentation allows parallel processing during training and enables selecting appropriate models for different video content types, improving overall coding efficiency while maintaining high accuracy through specialized models.
Solution Approach 2:
The system adjusts model parameters and convergence criteria based on the specific characteristics of training data partitions. By changing parameters such as learning rates, iteration counts, and convergence thresholds, the system optimizes the balance between prediction accuracy and computational efficiency for different coding scenarios.
3Loss of information
If machine learning prediction coding models are used, then compression efficiency is improved, but loss of time increases due to training and processing requirements
Solution Approach 1:
Training data partitioning and model training are performed in advance before actual video coding. The time-consuming training process is separated from real-time coding operations, so that during video compression, only the inference phase is executed, significantly reducing the time loss during actual processing while maintaining improved compression efficiency.
Data Source
AI summary
Video coding may include generating, by a processor, a decoded frame by decoding a current frame from an encoded bitstream and outputting a reconstructed frame based on the decoded frame. Decoding includes identifying a current encoded block from the current frame, identifying a prediction coding model for the current block, wherein the prediction coding model is a machine learning prediction coding model from a plurality of machine learning prediction coding models, identifying reference values for decoding the current block based on the prediction coding model, obtaining prediction values based on the prediction coding model and the reference values, generating a decoded block corresponding to the current encoded block based on the prediction values, and including the decoded block in the decoded frame.