Compound Intra Prediction for Multi-Directional Video Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding techniques often fail to optimally predict motion in video blocks due to reliance on a single intra-prediction mode, leading to blocking artifacts and inefficient use of computing resources.
Innovation Solution
Implement compound intra prediction by selecting and combining multiple intra-prediction modes based on rate-distortion analysis to generate a compound prediction block, which accounts for motion in multiple directions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a single intra-prediction mode is used, then the encoding process is simple and fast, but blocking artifacts occur and prediction accuracy deteriorates
Solution Approach 1:
The patent divides the current block into multiple sub-blocks and applies different intra-prediction modes to each sub-block based on its local characteristics. This segmentation allows the encoder to capture directional variations within the block, improving prediction accuracy while avoiding the need to apply a single complex mode to the entire block.
Solution Approach 2:
The patent dynamically selects intra-prediction modes for different sub-blocks based on rate-distortion analysis. Instead of using a static single mode for the entire block, the encoder evaluates multiple candidate modes and assigns the optimal mode to each sub-block, adapting to local image characteristics and motion patterns.
2Manufacturing precision
If multiple intra-prediction modes are combined, then prediction accuracy improves and blocking artifacts reduce, but computing resources increase
Solution Approach 1:
The patent applies multiple intra-prediction modes only to specific sub-blocks where they provide benefit, rather than uniformly applying them to the entire block. By performing rate-distortion analysis on a per-sub-block basis and selectively applying compound prediction only where needed, the encoder reduces unnecessary computing operations while maintaining prediction accuracy improvements.
3Object-generated harmful factors
If multiple intra-prediction modes are combined, then blocking artifacts are reduced, but the encoding process becomes more complex
Solution Approach 1:
The patent segments the block into sub-blocks and applies different prediction modes to each, which naturally reduces blocking artifacts at sub-block boundaries. This segmentation approach simplifies the complexity management by localizing the compound prediction application to specific regions rather than requiring complex global optimization.
Solution Approach 2:
The patent applies different intra-prediction modes to different sub-blocks based on their local characteristics such as edge orientations and texture patterns. This local quality approach ensures that each sub-block is predicted using the most appropriate mode for its specific characteristics, reducing blocking artifacts while keeping the overall encoding process manageable through localized decisions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Compound intra prediction is used to encode or decode blocks of video frames. First and second intra-prediction modes are selected based on motion within the video frame. For example, rate-distortion values resulting from predicting the motion can be determined for combinations of intra-prediction modes. The combination including the first and second intra-prediction modes can be selected based on it resulting in the lowest rate-distortion value. A compound prediction block is generated by combining first and second prediction blocks respectively generated using the first and second intra-prediction modes. For example, combining the first and second prediction blocks can include weighting the pixel values of the first and second prediction blocks or using each of those intra-prediction modes with different partitions of the block to be encoded or decoded. That block is then encoded or decoded using the compound prediction block.