Triangle Prediction Units for Efficient Complex-Scene Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently encoding and decoding high-definition and ultra-high-definition video data while maintaining image quality, particularly in handling geometrically complex video frames.
Innovation Solution
Implementing geometric prediction units, such as triangular prediction units, to partition video frames into multiple prediction units, constructing a merge list of motion vectors, and selecting appropriate motion vectors for motion compensation to enhance encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If block-based prediction units are used for video coding, then the coding process is simple and well-established, but the encoding efficiency for geometrically complex frames is insufficient
Solution Approach 1:
The coding block is divided into multiple prediction units with different geometric shapes (triangular, quadrangular, etc.) based on the actual content characteristics. This segmentation allows each prediction unit to be optimized for its specific geometric requirements, improving encoding efficiency for complex scenes while maintaining simplicity for regular patterns.
Solution Approach 2:
The prediction unit geometry is made dynamic rather than fixed. The system can adaptively select different geometric shapes (triangular, quadrangular, or other polygonal shapes) for prediction units based on the local image characteristics and motion patterns, enabling efficient coding of geometrically complex frames while maintaining backward compatibility with simpler block-based approaches.
2Manufacturing precision
If traditional rectangular block partitioning is used, then the implementation is straightforward, but image quality degradation occurs in geometrically complex regions
Solution Approach 1:
Different geometric shapes are applied to different regions of the image based on their local characteristics. Regions with geometric complexity use triangular or polygonal prediction units, while regions with regular patterns use traditional rectangular blocks. This local adaptation improves image quality in complex regions without unnecessarily complicating the entire coding system.
Solution Approach 2:
The patent introduces asymmetric geometric shapes (triangles, quadrangles) alongside the symmetric rectangular blocks. This asymmetry in prediction unit geometry allows the system to better fit irregular shapes in the image data, improving reconstruction quality in geometrically complex regions while maintaining the simplicity and symmetry of traditional rectangular blocks where appropriate.
3Productivity
If geometric prediction units are introduced to improve encoding efficiency, then coding performance increases, but the complexity of motion vector construction and selection increases
Solution Approach 1:
The motion vector construction and selection processes for geometric prediction units are designed to be universal and compatible with existing block-based motion estimation frameworks. The same basic principles of motion vector derivation, merge candidate list construction, and mode selection are applied, with only minor adaptations for geometric shapes. This universality improves coding efficiency while avoiding a complete redesign of the motion processing system.
Solution Approach 2:
The geometric prediction unit system is nested within the existing block-based video coding framework. Geometric prediction units can be applied at different levels of the coding tree structure, and the motion vector processing for geometric units builds upon the same foundations as traditional block-based units. This nested approach allows gradual integration of geometric prediction capabilities without overwhelming complexity.
Data Source
AI summary
Methods are provided for video coding. The method includes: partitioning video pictures into a plurality of coding units (CUs), at least one of which is further partitioned into two prediction units (PUs) including at least one geometric shaped PU; obtaining a first merge list including a plurality of candidates, each including one or more motion vectors; and obtaining a uni-prediction merge list for the geometric shaped PU by selecting the one or more motion vectors directly from the first merge list.