Intra Sub-Partition Video Coding With Template-Based Mode Derivation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face inefficiencies in compressing digital video signals, particularly in handling sub-partitions of coding blocks, leading to sub-optimal use of reference samples and prediction modes.
Innovation Solution
Combining template-based intra mode derivation (TIMD) and intra sub-partition (ISP) processes to utilize reconstructed samples from ISP for TIMD, allowing independent sub-partition level encoding/decoding and reordering based on identified conditions, especially for thin and small sub-partitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template-based intra mode derivation (TIMD) and intra sub-partition (ISP) processes are combined to use reconstructed samples from ISP in TIMD, then compression efficiency is improved, but processing complexity increases
Solution Approach 1:
The patent divides the coding block into multiple sub-partitions and applies independent TIMD processing to each sub-partition. This segmentation allows the system to handle smaller, more manageable units with their own template samples and prediction modes, improving overall compression efficiency while maintaining controlled processing complexity through localized operations.
Solution Approach 2:
The patent performs ISP processing first to generate reconstructed samples, which are then used as template samples for subsequent TIMD processing. This preliminary action of creating reconstructed samples beforehand enables the TIMD process to work with higher-quality reference data, improving compression efficiency without requiring simultaneous complex operations.
2Measurement precision
If independent sub-partition level TIMD derivation is performed for each sub-partition, then prediction accuracy is improved, but computational load increases
Solution Approach 1:
The patent applies different prediction modes and processing approaches to different sub-partitions based on their local characteristics. Each sub-partition undergoes independent TIMD derivation using its own template samples, allowing the system to optimize prediction accuracy for each local region while managing computational load through targeted processing only where needed.
Solution Approach 2:
The patent applies TIMD processing to all sub-partitions regardless of size, even though this increases computational load. This partial or excessive action ensures that prediction accuracy is maximized across the entire coding block by consistently applying the refined TIMD process to each sub-partition's template samples.
3Productivity
If reordering of encoding/decoding is implemented for thin and small sub-partitions, then use of reference samples is optimized, but processing complexity increases
Solution Approach 1:
The patent implements dynamic reordering of encoding/decoding operations based on the characteristics of sub-partitions. For thin and small sub-partitions, the processing order is adjusted to optimize reference sample availability, allowing the system to adaptively manage processing complexity while maximizing the utility of reference samples for prediction.
Data Source
AI summary
A coding block with ISP enabled may include multiple sub-partitions. For a first sub-partition in the coding block, an intra prediction mode may be derived based on the template samples associated with the first sub-partition. The first sub-partition may be decoded based on the intra prediction mode. For a second sub-partition in the coding block, an intra prediction mode may be derived based on the template samples associated with the second sub-partition, and the second sub-partition may be decoded based on that intra prediction mode. The sub-partitions in the coding block may independently derive an intra prediction mode using decoder side intra mode derivation and/or template-based intra mode derivation.