Residual Quadtree Transform Units for Rectangular Prediction Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing High Efficiency Video Coding (HEVC) standard faces challenges in applying hierarchical transform units (TUs) to residual blocks, particularly with rectangular and narrow rectangular Prediction Units (PUs), as the current methods complicate the transformation and prediction processes and are not compatible with the mandated transform tree structure.
Innovation Solution
A method is introduced to generate a transform tree using split-flags, allowing for the splitting and merging of TUs, enabling the application of square and rectangular TUs on rectangular PUs while maintaining the hierarchical transform tree structure, by parsing bitstreams to determine split-flags and modifying the transform tree accordingly, thereby merging multiple small TUs into larger ones and adjusting the tree structure to accommodate rectangular PUs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If narrow rectangular PUs (e.g., 1xN, 2xN) are used to match object boundaries, then prediction accuracy is improved, but the transform process becomes complicated and incompatible with the mandated square transform tree structure
Solution Approach 1:
The patent segments the transform process by introducing a separate transform tree generation step that operates independently from the prediction unit structure. The transform tree is constructed by splitting CUs based on split-flags from the bitstream, allowing TUs to be square (4x4 to 32x32) while PUs can remain narrow rectangular. This segmentation decouples the transform structure from the prediction structure, resolving the incompatibility between narrow PUs and square transform trees.
2Area of stationary object
If multiple small TUs are used to cover narrow rectangular PUs, then complete coverage is achieved, but the number of TUs increases and processing efficiency decreases
Solution Approach 1:
The patent merges multiple small square TUs that would otherwise be required to cover narrow rectangular PUs. By allowing the transform tree to be generated independently with square TUs of various sizes (4x4 to 32x32), adjacent small TUs can be merged into larger square TUs where appropriate. This merging reduces the total number of TUs while maintaining complete coverage of the PU area, thereby improving processing efficiency.
3Reliability
If the transform tree structure is strictly maintained with square TUs, then standard compatibility is improved, but the ability to efficiently represent rectangular PUs is worsened
Solution Approach 1:
The patent makes the transform tree generation process universal by introducing a configurable split-flag mechanism that can adapt to different PU shapes and sizes. The same square TU-based transform tree structure is used for all PUs, but the splitting behavior is controlled by split-flags parsed from the bitstream. This allows the universal square TU structure to efficiently represent both square and narrow rectangular PUs, maintaining standard compatibility while improving adaptability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A bitstream includes coded pictures, and split-flags for generating a transform tree. The bit stream is a partitioning of coding units (CUs) into Prediction Units (PUs). The transform tree is generated according to the split-flags. Nodes in the transform tree represent transform units (TU) associated with the CUs. The generation splits each TU only if the corresponding split-flag is set. For each PU that includes multiple TUs, the multiple TUs are merged into a larger TU, and the transform tree is modified according to the splitting and merging. Then, data contained in each PU can be decoded using the TUs associated with the PU according to the transform tree.