Constrained Multi-Type-Tree Block Partition for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding standards, such as HEVC, face challenges in efficiently partitioning blocks for prediction and transform, leading to increased coding complexity and signaling overhead, particularly with the Multi-Type-Tree (MTT) block partitioning method.
Innovation Solution
The proposed solution involves constrained block partitioning by prohibiting certain partition types and directions based on the partition type and direction of the previous partition level and the position of the current block, thereby reducing signaling overhead and improving performance. This is achieved by inferring the partition type or direction for splitting sub-blocks, such as prohibiting vertical binary tree partitioning for middle sub-blocks split by vertical triple tree partitioning, and similar constraints for other partition combinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Multi-Type-Tree (MTT) block partitioning method is used to improve coding efficiency, then prediction accuracy is improved, but signaling overhead and coding complexity increase
Solution Approach 1:
The patent changes the partitioning parameters by prohibiting certain partition types (binary tree and triple tree) for middle sub-blocks that are already split by triple tree partitioning. This constraint reduces the number of possible partition configurations from the full MTT set, thereby reducing signaling overhead while maintaining adequate prediction accuracy for regions already finely partitioned by triple tree splitting
Solution Approach 2:
The patent segments the block partitioning process into different rules based on sub-block position. Middle sub-blocks split by triple tree partitioning are prohibited from further binary or triple tree partitioning, while other sub-blocks retain full partitioning flexibility. This segmented approach reduces overall signaling complexity while preserving prediction accuracy in critical regions
2Adaptability or versatility
If MTT partitioning allows all partition types (binary tree and triple tree) for all sub-blocks, then adaptability is improved, but coding complexity increases
Solution Approach 1:
The patent applies different partitioning rules to different spatial locations within the block structure. Middle sub-blocks that are split by triple tree partitioning are subject to a prohibition on further binary or triple tree partitioning, while other sub-blocks maintain full partitioning flexibility. This local differentiation reduces coding complexity in regions where triple tree partitioning has already provided sufficient detail
Solution Approach 2:
The patent dynamically adjusts the available partitioning options based on the partitioning history and current block configuration. The prohibition rules are applied conditionally based on whether a parent block was split by triple tree partitioning and whether the current block is a middle sub-block. This dynamic constraint reduces the search space for optimal partitioning while maintaining adaptability where needed
3Reliability
If redundant partitioning is allowed in MTT method, then partitioning completeness is improved, but processing time increases
Solution Approach 1:
The patent preemptively prohibits certain partition types for middle sub-blocks split by triple tree partitioning, preventing redundant partitioning before it occurs. By anticipating that further binary or triple tree partitioning of these specific blocks would be redundant, the patent eliminates unnecessary processing steps while maintaining sufficient partitioning completeness through the initial triple tree split
Data Source
AI summary
A video coding system receives input data of a current block split from a parent block, and determines if the parent block is split by horizontal triple tree (TT) partitioning or vertical TT partitioning. A partition type selecting from TT partitioning and binary tree (BT) partitioning and a partition direction selecting from horizontal partitioning and vertical partitioning for splitting the current block are determined by prohibiting a horizontal partitioning type if the parent block is split by horizontal TT partitioning, and prohibiting a vertical partitioning type if the parent block is split by vertical TT partitioning, where the horizontal partitioning type and the vertical partitioning type depend on whether the current block is a middle sub-block in the parent block. The video coding system processes the current block according to the partition type and the partition direction.