Neural Network Video Prediction for UHD Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image encoding/decoding technologies face challenges in accurately predicting pixel values for high-resolution and high-definition images, particularly in handling inter and intra prediction for UHD images, which affects encoding efficiency and quality.
Innovation Solution
A prediction network is employed for encoding and decoding, utilizing intra and inter prediction networks to generate prediction blocks for target blocks, with preprocessing and postprocessing techniques like mean subtraction and normalization, and online parameter updates during decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional inter-prediction and intra-prediction technologies are used for high-resolution images, then encoding complexity increases, but prediction accuracy deteriorates
Solution Approach 1:
The patent replaces traditional mechanical prediction methods (inter-prediction and intra-prediction algorithms) with a neural network-based prediction system. The neural network learns optimal prediction strategies from training data and automatically generates prediction blocks, substituting complex manual algorithm design with automated machine learning-based prediction that achieves higher accuracy without proportional increases in encoding complexity.
2Measurement precision
If multiple prediction networks are used for different color channels, block sizes, and quantization parameters, then prediction accuracy improves, but device complexity increases
Solution Approach 1:
The patent segments the prediction task by creating specialized prediction networks for different color channels (Y, Cb, Cr), block sizes, and quantization parameter ranges. Each segmented network is trained on specific data characteristics, allowing them to specialize in particular prediction scenarios. This segmentation improves overall prediction accuracy by addressing the diverse characteristics of different image regions and encoding conditions with tailored prediction models.
Solution Approach 2:
The patent designs a universal prediction network architecture that can handle multiple color channels, block sizes, and quantization parameters through a unified framework. The network uses shared layers for common feature extraction and specialized branches for channel-specific processing, enabling one system to perform multiple prediction functions efficiently without requiring completely separate networks for each scenario.
3Measurement precision
If preprocessing and postprocessing are applied to reference blocks and prediction blocks, then prediction accuracy improves, but processing time increases
Solution Approach 1:
The patent applies preprocessing operations (such as normalization and color space transformation) to reference blocks before they are input to the prediction network. This preliminary preparation of data ensures that the neural network receives optimally formatted input, improving prediction accuracy by presenting data in the most effective representation for learning patterns without requiring complex processing during the actual prediction phase.
Data Source
AI summary
Disclosed herein are a method and apparatus for video decoding and a method and apparatus for video encoding. A prediction block for a target block is generated by predicting the target block using a prediction network, and a reconstructed block for the target block is generated based on the prediction block and a reconstructed residual block. The prediction network includes an intra-prediction network and an inter-prediction network and uses a spatial reference block and/or a temporal reference block when it performs prediction. For learning in the prediction network, a loss function is defined, and learning in the prediction network is performed based on the loss function.