Coding Block Partitioning With Context-Adaptive Split Direction Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as HEVC and VVC, face inefficiencies in block partitioning due to the use of fixed square blocks and lack of adaptive partitioning strategies, particularly in handling non-square block shapes, which affect compression efficiency.
Innovation Solution
Implementing a recursive multi-tree splitting method for partitioning pictures into coding blocks based on splitting information signaled in a data stream, using split, quad, and direction flags to determine block splits and suppress unnecessary flag decoding, thereby optimizing partitioning and reducing data stream size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If fixed square blocks are used for partitioning, then the partitioning structure is simple, but compression efficiency deteriorates due to inability to adapt to video signal characteristics
Solution Approach 1:
The picture is segmented into multiple types of blocks (coding blocks, transform blocks, prediction blocks) with different functions and sizes. The segmentation is performed recursively using multiple tree structures (quadtree, binary tree, ternary tree) allowing flexible division into square and non-square blocks, thereby improving compression efficiency while maintaining manageable structural complexity
Solution Approach 2:
The partitioning structure transitions from fixed to dynamic through the use of multiple tree types. The decoder adaptively selects among quadtree, binary tree, and ternary tree structures based on signaling information, enabling the partitioning to dynamically adapt to local video characteristics and improve compression performance
2Adaptability or versatility
If multiple tree structures are introduced for non-square blocks, then adaptability improves, but device complexity increases due to multiple partitioning trees
Solution Approach 1:
The solution adds a new dimension to the partitioning system by introducing multiple tree types (quadtree, binary tree, ternary tree) operating at different levels. This multi-dimensional approach allows the system to handle both square and non-square blocks efficiently, improving adaptability while the hierarchical organization keeps complexity manageable
Solution Approach 2:
The multiple tree structures serve universal purposes: quadtree for initial square block division, binary tree for non-square partitioning, and ternary tree for specific geometric patterns. This multi-functionality allows a single partitioning system to handle diverse block shapes and video characteristics without requiring separate specialized structures
3Loss of information
If all splitting flags are decoded from data stream, then partitioning information is complete, but data stream size increases
Solution Approach 1:
The solution extracts and transmits only the essential partitioning information in the data stream. By using compact flag structures and context-adaptive decoding, the system extracts the minimum necessary information to reconstruct the partitioning tree, reducing data stream size while maintaining complete partitioning information
Solution Approach 2:
Context-adaptive entropy decoding uses feedback from previously decoded information to optimize the decoding of splitting flags. The decoder adjusts probability models based on neighboring block patterns and previously decoded partitioning decisions, improving decoding efficiency and reducing the bit cost of transmitting partitioning information
Data Source
AI summary
An apparatus for decoding a picture is configured to partition the picture into coding blocks, using recursive multi-tree splitting. Splitting schemes for the multi-tree splitting are described. According to an embodiment, the decoder decodes a split direction flag from a data stream by context adaptive entropy decoding using a context which depends on whether, and discriminates between, a first number of criteria which would be fulfilled in case of a split direction indicated by the split direction flag being horizontal equals a second number of criteria which would be fulfilled in case of the split direction indicated by the split direction flag being vertical, the first number being greater than the second number, and the first number being smaller than the second number.