Asymmetric Sub-Block Prediction for Flexible Video Block Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video encoding/decoding methods are inefficient due to limited consideration of local characteristics within a video, as they primarily use square or rectangular blocks for partitioning, lacking flexibility in adapting to diverse image structures.
Innovation Solution
The method employs asymmetric sub-block partition structures, including quad-tree and binary-tree combinations, allowing for independent prediction and motion information storage across various asymmetric sub-blocks such as triangles, rectangles, and trapezoids, enhancing encoding/decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If square or rectangular blocks are used for partitioning, then the encoding/decoding process is simple, but the ability to adapt to diverse image structures is limited
Solution Approach 1:
The current block is divided into multiple asymmetric sub-blocks with different shapes (triangles, trapezoids, rectangles) rather than uniform square/rectangular partitions. This segmentation allows each sub-block to be optimized for specific local characteristics within the image, improving adaptability to diverse image structures while maintaining manageable complexity through systematic partitioning rules
Solution Approach 2:
The patent introduces asymmetric partition structures where sub-blocks have unequal sizes and varied geometries (triangular, trapezoidal, rectangular) instead of symmetric square or rectangular divisions. This asymmetry enables better fitting to local image features such as edges, textures, and motion patterns, directly addressing the need for diverse image structure adaptability
2Productivity
If independent prediction is performed for each sub-block, then the encoding efficiency is improved, but the computational complexity increases
Solution Approach 1:
Independent prediction is performed for each asymmetric sub-block, allowing different prediction modes, motion vectors, and prediction parameters to be applied locally to each sub-block based on its specific characteristics. This local quality approach improves encoding efficiency by capturing local variations in image content, while the use of standardized prediction algorithms keeps computational complexity manageable
Solution Approach 2:
The patent enables dynamic selection of prediction modes and parameters for each sub-block based on local image characteristics such as motion patterns, texture complexity, and edge orientations. This dynamic adaptation improves encoding efficiency by optimizing prediction accuracy for each region, while the systematic framework for mode selection prevents exponential growth in computational complexity
Data Source
AI summary
Disclosed herein are a video encoding/decoding method and apparatus. The video decoding method of the present disclosure includes: obtaining block partitioning information of a current block; based on the block partitioning information, partitioning the current block into a first sub-block and a second sub-block; deriving motion information of the first sub-block and motion information of the second sub-block respectively; based on the motion information of the first sub-block and the motion information of the second sub-block, generating a prediction sample of the first sub-block and a prediction sample of the second sub-block respectively; and through a weighted-sum of the prediction sample of the first sub-block and the prediction sample of the second sub-block, generating a prediction sample of the current block. Herein, the block partitioning information is index information indicating an index of a table including information indicating a plurality of predefined asymmetric partition shapes.