Symmetrical Transform Block Partitioning for Video Encoding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In next-generation video codecs following H.265/HEVC, the determination of orthogonal transformation types based on positional relations between prediction blocks and transform blocks is not feasible, leading to reduced encoding efficiency and increased calculation load due to the inability to perform block partitioning at orthogonal transformation.
Innovation Solution
The method involves partitioning prediction blocks into symmetrical rectangular transform blocks and determining orthogonal transformation types for each block based on its positional relation to the block center, using different types such as DCT-II, DST-VII, and DCT-VIII, to match the characteristics of the prediction residual signal, thereby improving encoding efficiency and reducing calculation load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If block partitioning is performed to determine orthogonal transformation types based on positional relations, then encoding efficiency is improved, but device complexity increases due to additional partitioning operations
Solution Approach 1:
The prediction block is divided into multiple transform blocks through systematic partitioning methods (e.g., 2×2, 4×4, or 8×8 subdivisions). This segmentation enables different orthogonal transformation types to be applied to different transform blocks based on their positional characteristics relative to the block center, thereby improving encoding efficiency without requiring complex adaptive decision-making at each block level.
Solution Approach 2:
Different orthogonal transformation types (such as DST-VII, DCT-II, or DCT-VIII) are selectively applied to different transform blocks based on their local characteristics. Transform blocks closer to the block center may use one transformation type while those at corners or edges use another, optimizing the transformation choice for each local region's signal characteristics.
2Productivity
If multiple orthogonal transformation types are used for different blocks, then encoding efficiency is improved, but calculation load increases
Solution Approach 1:
By dividing the block into a regular grid of transform blocks with predetermined partitioning patterns, the system avoids complex adaptive segmentation algorithms. Each transform block's position can be determined through simple coordinate calculations, reducing the computational overhead of block division while enabling multiple transformation types to be applied efficiently.
Solution Approach 2:
The patent changes the transformation type parameter based on the transform block's position relative to the block center. Instead of performing complex analysis to determine optimal transformations, the system uses position-based parameter selection (e.g., corner blocks use one type, edge blocks use another, center blocks use a third), significantly reducing calculation load while maintaining encoding efficiency.
3Ease of operation
If symmetrical partitioning is used for transform blocks, then ease of operation is improved, but adaptability decreases due to restricted partitioning patterns
Solution Approach 1:
The symmetrical partitioning scheme serves multiple functions: it simplifies the determination of transform block positions, enables consistent application of position-based transformation type selection, and works effectively for blocks of various sizes. The same partitioning logic can be applied universally across different block dimensions, providing both operational simplicity and broad adaptability.
Solution Approach 2:
While the overall partitioning pattern is symmetrical, the patent accommodates asymmetric needs by allowing different orthogonal transformation types to be applied to symmetrically positioned blocks based on their relationship to the center. This enables asymmetric transformation selection (different types for different regions) while maintaining symmetric partitioning structure, balancing ease of operation with adaptability.
Data Source
AI summary
An encoding method for encoding an image using an inter-picture prediction includes determining a prediction block on which the inter-picture prediction is to be performed, partitioning the prediction block into a plurality of transform blocks by a partitioning method that partitions the prediction block, so that boundaries of the plurality of transform blocks are symmetrical with respect to a horizontal line passing a center of the prediction block and are symmetrical with respect to a vertical line passing the center of the prediction block, the plurality of transform blocks being rectangular, and determining, for each of the plurality of transform blocks, an orthogonal transformation type used for each of a vertical direction and a horizontal direction of a given transform block of the plurality of transform blocks based on a positional relation between the given transform block and the center of the prediction block.