Multi-type-tree video coding framework for asymmetric block partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as HEVC, lack flexibility in block partitioning, particularly when objects are centered within blocks, leading to inefficient coding performance due to limited partitioning modes that often split blocks symmetrically or not at all.
Innovation Solution
The implementation of a multi-type-tree (MTT) framework that allows for multiple partitioning structures at each depth of a tree structure, including quad-tree, binary-tree, and triple-tree types, enabling asymmetric and non-central splits to better capture objects within blocks, thereby improving coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional quad-tree-binary-tree partitioning is used, then the block partitioning structure is simple and easy to implement, but the coding efficiency deteriorates when objects are centrally located within blocks
Solution Approach 1:
The patent extends the traditional binary tree segmentation into a multi-type tree structure that includes quad-tree, binary-tree, and triple-tree partitioning modes. This allows the block to be divided into different numbers of sub-blocks (4, 2, or 3) depending on the partitioning type, providing more flexible segmentation that can adapt to various object positions and shapes within the block.
Solution Approach 2:
The patent introduces dynamic selection of partitioning types at each node of the tree structure. Instead of using a fixed partitioning mode, the system can dynamically choose from multiple partitioning types (quad-tree, binary-tree, triple-tree) based on the characteristics of the block being processed, allowing the partitioning structure to adapt to different coding scenarios.
2Adaptability or versatility
If symmetric block splitting is used, then the partitioning implementation is straightforward, but the ability to capture centrally located objects deteriorates
Solution Approach 1:
The patent introduces asymmetric partitioning modes that allow blocks to be divided into sub-blocks of different sizes and positions. Specifically, the triple-tree partitioning mode enables asymmetric division where the center of the block does not need to be split, allowing asymmetric arrangement of sub-blocks that can better capture objects located at the center or in specific regions of the block.
Solution Approach 2:
The patent allows different partitioning types to be applied to different regions or nodes in the tree structure based on local characteristics. This means that some blocks may use quad-tree partitioning while others use binary or triple-tree partitioning, depending on whether they contain centrally located objects or not, optimizing the partitioning strategy for each local region.
3Productivity
If limited partitioning modes are used, then the system complexity is low, but the compression performance deteriorates
Solution Approach 1:
The patent creates a universal multi-type tree framework that incorporates multiple partitioning modes (quad-tree, binary-tree, triple-tree) within a single unified structure. This allows the system to handle various coding scenarios using a common framework, where each partitioning type can be selected based on the specific needs of the block, providing multi-functionality that improves compression performance without requiring separate systems for each partitioning approach.
Data Source
AI summary
A method of decoding video comprising: receiving an encoded block of video data, determining a transform for the encoded block of video data, wherein the transform has a size S that is not a power of two, rounding S to a power of two creating a transform with a modified size S′, applying an inverse transform with the modified size S′ to the encoded block of video data to create residual video data, and decoding the residual video data to create decoded block of video data.