Video Encoder Differentiable Prediction Motion Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in accurately estimating motion beyond a certain level, especially as video resolution increases, due to limited search ranges in mobile devices and software optimizations.
Innovation Solution
A video encoder is designed with a differentiable prediction (DP) module that performs a full search in a predetermined area to output an optimal initial search position, and a motion estimation (ME) module that moves the search position towards this optimal initial search position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a maximum search range is set within the system-on-chip (SoC) or a search range is set for software optimization, then device complexity and processing efficiency are improved, but motion estimation precision deteriorates for high-resolution videos
Solution Approach 1:
The patent applies preliminary action by using a neural network to predict the optimal initial search position before performing motion estimation. The neural network processes historical motion information and frame characteristics to pre-determine the most likely motion vector direction and magnitude, thereby guiding the subsequent search process and reducing the effective search range needed while maintaining high precision for high-resolution videos
Solution Approach 2:
The patent introduces a neural network as an intermediary component between the frame processing units and the motion estimation unit. This intermediary learns from historical data to provide optimized initial search positions, effectively mediating between the need for comprehensive search (precision) and limited search range (efficiency)
2Measurement precision
If the search range is increased to capture motion beyond a certain level, then motion estimation precision is improved, but device complexity and computational load increase
Solution Approach 1:
The neural network performs preliminary prediction of motion characteristics based on historical motion vectors and frame properties. This pre-computation of optimal search positions allows the system to focus computational resources only on the most promising search regions rather than exhaustively searching the entire possible motion range, thereby maintaining precision while reducing complexity
Solution Approach 2:
The patent applies local quality by adapting the search range and strategy to local motion characteristics. Instead of using a uniform large search range for the entire frame, the system adjusts the search parameters locally based on predicted motion patterns, achieving high precision where needed while avoiding unnecessary computational complexity in regions with small or predictable motion
Data Source
AI summary
A video encoder is provided. The video encoder according to an example embodiment includes: a differentiable prediction (DP) module configured to output an optimal initial search position by performing full search in a predetermined area by using a pair of frames of a video as input; and a motion estimation (ME) module configured to perform motion estimation by moving a search position toward the optimal initial search position output by the DP module.