Video Encoding Using Machine Learning Categorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in encoding efficiency for high-resolution videos like 4K/8K, leading to increased encoder complexity and reduced decoding efficiency, particularly due to the need for signaling significant signal bits for mode decisions in adaptive filtering techniques.
Innovation Solution
The method involves extracting features from reconstructed pixel groups, categorizing them using a machine learning model, and selecting a parameter set for encoding/decoding processes, allowing for efficient encoding/decoding without explicitly signaling the parameter set in the bitstream, thereby reducing the code amount and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If adaptive filtering techniques and intra prediction techniques are used to improve encoding efficiency, then encoding efficiency is improved, but encoder complexity increases
Solution Approach 1:
The system uses machine learning models that automatically categorize reconstructed pixel groups and select parameter sets without requiring complex encoder decision-making processes. The ML models perform the adaptive selection self-service, reducing encoder complexity while maintaining encoding efficiency improvements from adaptive filtering and intra prediction techniques.
Solution Approach 2:
The invention changes the parameter selection approach by using machine learning-based categorization instead of traditional complex decision-making. Different parameter sets are selected based on ML-categorized pixel group characteristics, enabling adaptive filtering and intra prediction techniques to improve encoding efficiency without proportionally increasing encoder complexity.
2Productivity
If mode decisions are signaled at small unit level (e.g. 4×4 block) with many possible modes, then encoding efficiency is improved, but signal bit size increases
Solution Approach 1:
Instead of signaling all possible mode decisions at the smallest unit level, the system applies machine learning categorization to select only the necessary parameter sets. This partial action approach maintains encoding efficiency improvements while avoiding the excessive signal bit size that would result from comprehensive mode signaling at every small block level.
Solution Approach 2:
The machine learning model acts as an intermediary between the reconstructed pixel groups and the parameter set selection. Rather than directly signaling complex mode decisions, the ML model categorizes pixel groups and selects appropriate parameter sets, reducing the signal bit size while preserving encoding efficiency benefits.
3Productivity
If adaptive tools are applied to improve encoding efficiency, then encoding efficiency is improved, but the number of signal bits required increases
Solution Approach 1:
The system performs preliminary machine learning-based categorization of reconstructed pixel groups before encoding. This preliminary action identifies and selects appropriate parameter sets in advance, enabling adaptive tools to improve encoding efficiency without requiring excessive signal bits to be transmitted during the encoding process.
Data Source
AI summary
To provide an encoding method for reducing deterioration in encoding efficiency when encoding a video/an image by selecting one operation mode out of a number of operation modes. The encoding method includes: extracting one or more features from one or more reconstructed pixel groups; categorizing the one or more reconstructed pixel groups using the extracted one or more features and a first machine learning model; and performing an encoding process on a current block of image samples included in a current picture using a parameter set corresponding to a result of the categorizing, the parameter set being one of a plurality of predefined parameter sets.