Asymmetric Binary Partitioning for Video Block Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The HEVC video compression standard lacks flexibility in coding unit representation, particularly for asymmetric binary partitioning, which hampers the trade-off between encoder and decoder complexity and compression efficiency.
Innovation Solution
The method involves partitioning blocks into asymmetric binary tree units with sizes not limited to powers of 2, allowing for residual splitting and transformation into transform coefficients, followed by quantization and entropy coding, and includes adaptive use of asymmetric binary partitioning based on block sizes and usage rates to optimize compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If asymmetric binary partitioning is used to improve compression efficiency, then compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent applies asymmetric binary partitioning by dividing coding units into non-square blocks with different width and height ratios (e.g., 1:2, 2:1, 1:3, 3:1). This allows the encoder and decoder to process image blocks that better match the actual content geometry, improving compression efficiency while maintaining manageable complexity through standardized partitioning rules.
Solution Approach 2:
The patent introduces dynamic block size and shape selection during the encoding process. The encoder can adaptively choose from multiple partitioning modes (horizontal split, vertical split, different split positions) based on the local image characteristics, allowing the system to optimize compression efficiency dynamically without requiring exhaustive complexity.
2Productivity
If flexible block partitioning is used to improve compression efficiency, then compression efficiency is improved, but encoding time increases
Solution Approach 1:
The patent applies partial action by implementing asymmetric binary partitioning selectively rather than universally. The encoder can choose to apply asymmetric partitioning only to coding units where it provides significant compression benefits, while using simpler symmetric partitioning elsewhere, thus improving compression efficiency without proportionally increasing encoding time across the entire image.
Solution Approach 2:
The patent changes the partitioning parameters (block size, block shape, split position) to optimize the trade-off between compression efficiency and encoding time. By adjusting these parameters adaptively based on image content characteristics, the system achieves better compression without linearly increasing encoding complexity.
Data Source
AI summary
Different implementations are described, particularly implementations for video encoding and decoding based on asymmetric binary partitioning of image blocks are presented. The encoding method comprises for a picture, wherein at least one component of the picture is divided into blocks of samples, partitioning a block into block partitions, wherein at least one block partition has a size equal to a positive integer different from a power of 2 in width and/or height, generating a residual by predicting the block partitions and subtracting the prediction from the block partitions, splitting the generated residual into at least two residual blocks with a size equal to a power of 2 in width and height and transforming the at least two residual blocks into transform coefficients, quantizing the transform coefficients, and entropy coding of the quantized transform coefficients. Correspondingly, the decoding method comprises receiving in a bitstream quantized transform coefficients, entropy decoding, de-quantizing and inverse transforming the transform coefficients to obtain residual blocks with a size equal to a power of 2 in width and height, merging at least two residual blocks to reconstruct a residual for a block partition having a size equal to a positive integer different from a power of 2 in width and/or height and reconstructing a block of the picture by predicting the block partitions and combining the residuals with the predicted block partitions. Ohers embodiments are presented for a block partitioning on the border of the picture and for setting maximum and/or minimum block sizes.