Intra Prediction Unit Classification for Faster Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding methods incur high computational complexity and low efficiency due to traversing all available intra-frame prediction modes for each prediction unit, leading to suboptimal video encoding performance.
Innovation Solution
Determine the intra-frame prediction mode of a prediction unit by analyzing its distribution information and categorizing it based on the texture characteristics of neighboring units, thereby reducing the need to traverse all prediction modes and optimizing the encoding process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all available intra-frame prediction modes are traversed for each prediction unit to determine the optimal mode, then the prediction mode determination is thorough and accurate, but the computing amount increases significantly and encoding efficiency decreases
Solution Approach 1:
The patent applies preliminary action by pre-determining the unit category of each prediction unit based on its position and neighboring units before actual encoding. This classification is performed in advance using spatial relationship analysis, so that when encoding occurs, the system already knows which prediction modes are most likely to be optimal, allowing it to skip traversing all modes and directly select from a reduced set, thereby significantly reducing computational complexity while maintaining accuracy
Solution Approach 2:
The patent applies local quality by analyzing the specific characteristics of each prediction unit's location and its neighboring units to determine unit categories. Different spatial locations and neighboring configurations result in different categories, allowing the system to adapt the prediction mode selection strategy to local characteristics. This enables optimized prediction mode determination tailored to each specific PU context rather than using a uniform approach, improving both accuracy and efficiency
2Reliability
If all available intra-frame prediction modes are traversed for each prediction unit, then the encoding process is comprehensive, but the computing resources are consumed excessively
Solution Approach 1:
The system performs preliminary classification of prediction units into categories based on spatial relationships and neighboring unit characteristics before the actual encoding process. This pre-computation of unit categories reduces the search space for prediction modes during encoding, allowing the system to maintain comprehensive encoding coverage while significantly reducing computing resource consumption by avoiding unnecessary traversal of all prediction modes
Solution Approach 2:
Instead of traversing all possible prediction modes (excessive action), the patent uses partial action by determining unit categories that enable selection from a subset of relevant prediction modes. This partial traversal approach maintains sufficient encoding reliability by focusing computational effort only on the most likely prediction modes for each unit category, thereby reducing overall computing resource consumption while preserving encoding quality
Data Source
AI summary
Disclosed are a video encoding method performed by an electronic device. The method includes: determining prediction unit (PU) distribution information of an image block in which a current PU is located; determining a unit category of a current PU according to the PU distribution information of the image block; determining an intra-frame prediction mode of the current PU according to the unit category of the current PU; and encoding the image block using the current PU according to the determined intra-frame prediction mode. According to the embodiments of this application, the intra-frame prediction mode of the PU can be rapidly determined, to improve video encoding efficiency.