Video Decoder Transform Signaling Overhead Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video encoding and decoding, the existing methods face inefficiencies due to the need to signal transform types for each transform unit, especially when multiple split types and transforms are used, leading to increased bit costs and complexity.
Innovation Solution
The proposed method restricts transform selection by determining transforms for transform units based on spatially neighboring units, using similar or complementary transforms along and perpendicular to splitting lines, reducing the need for explicit signaling of transform types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple transform types are signaled for each transform unit, then transform accuracy is improved, but bit cost increases
Solution Approach 1:
The patent merges transform type information from spatially neighboring transform units. Instead of signaling transform types independently for each TU, the decoder infers transform types for some TUs based on the transform types of neighboring TUs, combining information across multiple units to reduce redundancy.
Solution Approach 2:
The patent establishes universal transform type relationships that apply across different split types (quad-tree, binary-tree, ternary-tree). The same set of transform type combinations is used for all split types, creating a universal signaling scheme that works regardless of the specific partitioning structure.
2Adaptability or versatility
If multiple transform types are signaled for each transform unit, then coding flexibility is improved, but device complexity increases
Solution Approach 1:
The patent applies local quality by allowing different transform type relationships for different spatial relationships. Transform types are determined based on local neighborhood relationships (horizontal, vertical, diagonal neighbors), enabling adaptive transform selection that respects local image characteristics while following systematic rules.
Solution Approach 2:
The patent performs preliminary action by establishing transform type relationships in a predetermined order during decoding. The decoder systematically determines transform types for TUs in a specific sequence based on their spatial relationships to already-decoded neighboring TUs, avoiding the need to evaluate all possible combinations.
3Manufacturing precision
If transform types are independently determined for each transform unit, then transform optimization is improved, but signaling overhead increases
Solution Approach 1:
The patent segments the transform type determination process into independent rules for different spatial relationships. Each TU's transform type is determined by applying specific segmentation rules based on its relationship to neighboring TUs (e.g., horizontal neighbor relationship, vertical neighbor relationship), allowing modular and efficient processing.
Solution Approach 2:
The patent uses copying by having the decoder infer transform types for some TUs by copying transform type information from neighboring TUs. Instead of signaling all transform types explicitly, the decoder copies transform type decisions from already-decoded neighboring units, reducing the amount of information that needs to be transmitted.
Data Source
AI summary
A decoding method is presented. a type of split of a block into transform units is first decoded. A transform is then determined for each transform unit of said block responsive to said type of split. Finally, decoded transform coefficients of said transform units are inverse transformed using the determined transforms.