Significance Map Context Selection for Large Transform Units
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding standards, such as H.264/AVC and developing MPEG-H, face inefficiencies in encoding and decoding significance maps due to the high percentage of data occupied by encoding quantized transform coefficients, particularly in larger transform units, which requires managing numerous contexts for entropy encoding and decoding.
Innovation Solution
The implementation of multi-level significance maps and context-adaptive encoding/decoding methods, where significant-coefficient flags are processed in sets and context sets are selected based on position and neighboring flags, allowing for efficient encoding and decoding of significant-coefficient flags within larger transform units like 16x16 and 32x32 TUs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If context-adaptive encoding is used for significance maps in larger transform units, then compression performance is improved, but decoding complexity increases due to extensive context tracking and lookup
Solution Approach 1:
The patent divides the transform unit into multiple coefficient groups (e.g., 4x4 groups within a 16x16 or 32x32 TU). Each coefficient group is processed independently with its own context model, reducing the overall complexity by breaking down the large transform unit into smaller, manageable segments that can be decoded with fewer context lookups.
Solution Approach 2:
The patent applies different context models to different regions of the transform unit based on local characteristics. Specifically, it uses position-dependent context selection where the context set is chosen based on the coefficient's position within the transform unit and neighboring significant-coefficient flags, allowing the decoding complexity to be adapted locally rather than uniformly across the entire transform unit.
2Measurement precision
If extensive context tracking is used during entropy encoding, then encoding precision is improved, but processing time increases
Solution Approach 1:
The patent pre-defines multiple context sets with different context models before the actual encoding/decoding process. These context sets are prepared in advance and can be quickly selected based on position and neighboring flags, eliminating the need for complex real-time context tracking and lookup during the main processing loop, thus reducing processing time while maintaining encoding precision.
3Productivity
If multi-level significance maps are used for larger transform units, then compression efficiency is improved, but context selection complexity increases
Solution Approach 1:
The patent implements dynamic context set selection where the appropriate context set is chosen based on the current coefficient's position within the transform unit and the values of neighboring significant-coefficient flags. This dynamic adaptation allows the system to optimize compression efficiency for each specific position and pattern while managing complexity through rule-based selection rather than exhaustive search.
Solution Approach 2:
The patent creates multiple context sets that are essentially copies of the same context model structure, but each tailored for specific regions or patterns within the transform unit. By preparing these copied context sets in advance with slight variations optimized for different positions and neighboring flag patterns, the system achieves high compression efficiency without the computational burden of creating entirely new context models for each case.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods of encoding and decoding for video data are described for encoding or decoding multi-level significance maps. Distinct context sets may be used for encoding the significant-coefficient flags in different regions of the transform unit. In a fixed case, the regions are defined by coefficient group borders. In one example, the upper-left coefficient group is a first region and the other coefficient groups are a second region. In a dynamic case, the regions are defined by coefficient group borders, but the encoder and decoder dynamically determine in which region each coefficient group belongs. Coefficient groups may be assigned to one region or another based on, for example, whether their respective significant-coefficient-group flags were inferred or not.