Intra Prediction Fusion Using Principal and Fusion Modes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding/decoding technologies lack efficiency in intra prediction, particularly in the use of single mode prediction for blocks, limiting the effectiveness of fusion-based prediction modes.
Innovation Solution
Introduce a fusion intra prediction coding tool that allows multiple modes for intra-predicted blocks, using a principal mode and one or more fusion modes, with mode selection based on previously decoded blocks and spatial neighborhood analysis to optimize signaling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple fusion intra prediction modes are used for intra-predicted blocks, then prediction accuracy is improved, but signaling overhead increases
Solution Approach 1:
The patent applies local quality by differentiating the treatment of intra-predicted blocks based on their MPM syntax element status. Blocks with MPM signaled receive fusion prediction treatment while those without use standard prediction, optimizing the balance between prediction accuracy and signaling overhead for different block types locally
Solution Approach 2:
The patent implements partial action by applying fusion intra prediction modes selectively only to blocks where MPM syntax element is signaled, rather than universally to all intra-predicted blocks. This partial application reduces overall signaling overhead while still improving prediction accuracy for the subset of blocks that benefit most
2Productivity
If fusion intra prediction modes are applied to all intra-predicted blocks, then prediction efficiency is improved, but device complexity increases
Solution Approach 1:
The patent reduces device complexity by applying fusion prediction processing locally only to blocks with MPM syntax element signaled, rather than implementing complex fusion logic for all intra-predicted blocks. This selective approach maintains prediction efficiency for relevant blocks while simplifying the overall device complexity
3Measurement precision
If additional syntax elements are signaled for fusion modes, then prediction accuracy is improved, but data stream size increases
Solution Approach 1:
The patent applies partial action by signaling fusion mode syntax elements only for blocks where MPM is already signaled, rather than for all intra-predicted blocks. This selective signaling reduces the overall increase in data stream size while still achieving improved prediction accuracy for the targeted subset of blocks
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is a method for decoding a picture from a data stream, comprising determining, based on a most probable mode syntax element signaled in the data stream, a principal intra-prediction mode out of a first set of intra prediction modes, forming a fused prediction signal for an intra-predicted block by fusing a first prediction signal obtained using the principal intra prediction mode and one or more second prediction signals obtained using one or more fusion intra prediction modes; reconstructing the intra-predicted block using the fused prediction signal.