Wedgelet Partitioning for Video Block Motion Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video coding techniques, such as those under the H.264 standard, face inefficiencies due to the fixed square or rectangular macroblock and sub-macroblock shapes, which do not accurately match the outline of moving objects, leading to reduced coding efficiency when multiple regions within a macroblock exhibit different motion.
Innovation Solution
The method involves partitioning macroblocks into wedgelets, allowing for more flexible and accurate motion compensation by dividing blocks into wedges that can be oriented at any angle, enabling better representation of edges and reducing computational complexity through efficient search strategies and shape-adaptive transforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional quad-tree decomposition with fixed square or rectangular macroblocks is used, then the coding process is simple and computationally efficient, but the shape of macroblocks does not correspond to the outline of moving objects, reducing coding efficiency
Solution Approach 1:
The patent segments macroblocks into wedgelets by dividing them into two triangular regions separated by a wedgelet boundary line. This segmentation allows each triangle to be processed independently with potentially different motion vectors, improving the ability to represent objects with arbitrary shapes while maintaining a systematic decomposition approach similar to conventional quad-tree methods.
Solution Approach 2:
The patent introduces asymmetric wedgelet shapes (triangular regions) to replace the symmetric square/rectangular macroblocks. By allowing wedgelets to be formed with boundaries at various angles rather than only horizontal and vertical lines, the system can better match the asymmetric outlines of moving objects, thereby improving coding efficiency.
2Device complexity
If fixed square or rectangular macroblocks are used, then the decomposition structure is simple, but multiple regions with different motion cannot be efficiently coded within the same macroblock
Solution Approach 1:
The patent introduces dynamic wedgelet partitioning where the wedgelet boundary line can be positioned at different angles and locations within a macroblock based on the actual motion characteristics of different regions. This dynamic adaptation allows the decomposition structure to flexibly match the motion patterns in the video content, improving motion information accuracy without requiring a completely complex new structure.
Solution Approach 2:
The patent applies local quality by allowing different regions (triangular wedgelets) within a macroblock to have different properties, specifically different motion vectors and potentially different coding modes. This enables each local region to be optimized for its specific motion characteristics, improving overall motion compensation accuracy while maintaining a relatively simple overall structure.
3Manufacturing precision
If macroblocks are partitioned into wedgelets with arbitrary angles, then edge representation is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by selectively applying wedgelet partitioning only to macroblocks that benefit from it, rather than forcing wedgelet decomposition on all macroblocks. By using criteria such as motion vector differences or gradient analysis to identify candidates for wedgelet partitioning, the system achieves improved edge representation where needed while avoiding unnecessary computational complexity in regions where conventional macroblocks are sufficient.
Data Source
AI summary
In a method for decomposing a block of a video sequence frame, it is determined as to whether either or both of the dimensions of the block equals a predetermined minimum dimension. In response to either or both of the dimensions equaling the predetermined minimum dimension, a motion vector for the block is sent. In response to both of the dimensions exceeding the predetermined minimum dimension, a motion vector for the block is determined, the block is partitioned into two wedges, the block is divided into four N/2×N/2 sub-blocks, and these steps are repeated on each of the four N/2×N/2 sub-blocks until either or both of the dimensions equal the predetermined minimum dimension.