Constrained Coding Tree Partitioning for Faster Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently partitioning coding tree units (CTUs) without breaking pipeline structures, leading to coding slowdowns and inefficiencies.
Innovation Solution
Implementing a constrained coding tree that restricts split modes based on the relationship between coding tree node dimensions and maximum transform unit dimensions, preventing splits that cross pipeline block boundaries, thereby optimizing partitioning and encoding speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a coding tree node is split into smaller blocks, then the partitioning flexibility is improved, but the pipeline structure may be broken causing coding slowdowns
Solution Approach 1:
The patent segments the coding tree node splitting process into distinct cases based on node dimensions. By dividing the problem into specific dimension ranges (e.g., height >= 2*maxTUheight and width >= 2*maxTUwidth), the patent enables selective application of split modes that maintain pipeline structure while preserving partitioning flexibility where applicable.
Solution Approach 2:
The patent changes the parameters governing split mode selection by introducing dimension-based constraints. Instead of allowing all split modes uniformly, the patent modifies the selection criteria based on specific dimension relationships (e.g., comparing node height/width to multiples of maxTUheight/maxTUwidth), thereby controlling which splits are permitted to maintain pipeline integrity.
2Manufacturing precision
If more split modes are available for coding tree nodes, then the encoding precision is improved, but the device complexity increases
Solution Approach 1:
The patent applies local quality by making split mode availability dependent on local dimension characteristics of each coding tree node. Different regions of the coding tree (based on their dimensions relative to maxTU dimensions) have different sets of available split modes, optimizing precision where needed while reducing complexity where constraints apply.
Solution Approach 2:
The patent changes the parameter set for split mode selection by introducing dimension-based thresholds. The available split modes are determined by parameters such as whether height >= 2*maxTUheight and width >= 2*maxTUwidth, dynamically adjusting the complexity of split mode selection based on node characteristics.
3Area of stationary object
If the coding tree node dimensions are larger, then the transform unit coverage is improved, but the pipeline block boundary alignment becomes more difficult
Solution Approach 1:
The patent changes the dimensional parameters of coding tree nodes to be multiples of maxTU dimensions (e.g., height >= 2*maxTUheight, width >= 2*maxTUwidth). This parameterization ensures that nodes are sized to cover complete transform unit areas while maintaining alignment with pipeline block boundaries, as the dimensional relationships are explicitly defined in terms of maxTU multiples.
Data Source
AI summary
A video coding mechanism is disclosed. The mechanism includes partitioning an image to create at least one coding tree unit (CTU) with at least one coding tree node. The mechanism further includes determining that the coding tree node height is twice a maximum transform unit (TU) height and the coding tree node width is twice a maximum TU width. The mechanism further includes selecting a split mode for the coding tree node based on the determination. The split mode is selected from a quad-tree split, a horizontal binary-tree split, and no split. The mechanism further includes applying the split mode to the coding tree node to create one or more coding units (CUs). The mechanism further includes encoding the CUs into a bitstream. The mechanism further includes transmitting the bitstream toward a decoder.