ML Intra-Prediction Mode Selection for Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in achieving improved coding efficiencies while reducing computational complexity, as increased coding efficiency leads to higher computational demands.
Innovation Solution
The implementation of a machine-learning (ML) model-based method for encoding video blocks using intra-prediction, which involves obtaining ML intra-prediction modes, selecting an encoding intra-prediction mode based on the relative reliabilities of ML and most-probable modes, and encoding this mode in a compressed bitstream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If improved coding efficiency is achieved through traditional encoding techniques, then video quality is maintained or enhanced, but computational complexity increases
Solution Approach 1:
The patent replaces traditional mechanical encoding algorithms with a machine learning model that uses pre-calculated features and decision fusion to determine intra-prediction modes. This substitution reduces computational complexity during encoding while maintaining coding efficiency through the trained ML model's ability to make accurate predictions with fewer operations.
Solution Approach 2:
The patent performs preliminary calculations of features (such as gradient information, texture characteristics, and edge detection) before the actual encoding process. These pre-calculated features are stored and reused during mode decision-making, eliminating the need to recalculate them repeatedly and thereby reducing overall computational complexity while preserving coding efficiency.
2Productivity
If traditional intra-prediction modes are used for encoding, then computational complexity is manageable, but coding efficiency is limited
Solution Approach 1:
The patent implements a dynamic mode selection mechanism that adaptively chooses between ML-generated modes and traditional most-probable modes based on decision fusion logic. This dynamic approach allows the system to leverage ML capabilities for improved coding efficiency while falling back to simpler traditional methods when appropriate, thereby managing computational complexity effectively.
Solution Approach 2:
The patent applies different encoding strategies to different regions or blocks of video data based on local characteristics. By analyzing local features and applying ML-based mode selection only where beneficial while using traditional modes elsewhere, the system improves overall coding efficiency without uniformly increasing computational complexity across all video content.
Data Source
AI summary
Ultra light models and decision fusion for increasing the speed of intra-prediction are described. Using a machine-learning (ML) model, an ML intra-prediction mode is obtained. A most-probable intra-prediction mode is obtained from amongst available intra-prediction modes for encoding the current block. As an encoding intra-prediction mode, one of the ML intra-prediction mode or the most-probable intra-prediction mode is selected, and the encoding intra-prediction mode is encoded in a compressed bitstream. A current block is encoded using the encoding intra-prediction mode. Selection of the encoding intra-prediction mode is based on relative reliabilities of the ML intra-prediction mode and the most-probable intra-prediction mode.