Non-Square Transform Partitions in Intra-Prediction Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding techniques are limited in their ability to efficiently utilize non-square transform partitions, which can lead to suboptimal compression and quality in video encoding processes.
Innovation Solution
The proposed solution involves partitioning coding units into multiple prediction units and determining non-square transform partitions for each unit, with syntax elements signaling the type and direction of partitions, allowing for flexible and efficient use of non-square transforms in intra-prediction video coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If square transform partitions are used in video coding, then the encoding process is simple and standardized, but the compression efficiency and video quality are suboptimal
Solution Approach 1:
The transform unit is segmented into multiple sub-transform units arranged in a grid pattern, allowing different square transform sizes (e.g., 4x4, 8x8, 16x16) to be applied to different regions of the same block based on local characteristics, thus improving compression efficiency while maintaining standardized processing
Solution Approach 2:
The transform partition structure is made dynamic by allowing the encoder to select different square transform sizes for different sub-transform units within a block, adapting the transform granularity to the local frequency characteristics and content complexity of each region
2Productivity
If non-square transform partitions are introduced to improve compression efficiency, then video quality and compression performance are enhanced, but the encoding complexity and device complexity increase
Solution Approach 1:
Non-square transform partitions are introduced where transform units can have different widths and heights (e.g., 32x16, 16x32, 64x32, 32x64), allowing the transform shape to match the anisotropic characteristics of video content such as horizontal or vertical edges, thereby improving compression efficiency for directional features
Solution Approach 2:
Different transform partition shapes are applied to different regions of the video block based on local characteristics - for example, non-square partitions aligned with dominant edge directions are used in regions with strong directional content, while square partitions are used in isotropic regions, optimizing compression locally
3Adaptability or versatility
If multiple transform partition types are supported, then flexibility in encoding is improved, but the syntax complexity and signaling overhead increase
Solution Approach 1:
Instead of supporting all possible non-square partition combinations, the implementation supports a selected subset of non-square transform partitions (e.g., only horizontal and vertical orientations at specific size ratios), providing sufficient flexibility for most video content while limiting syntax complexity to manageable levels
Solution Approach 2:
The transform partition structure is extended from one-dimensional (single size) to two-dimensional (width and height dimensions), allowing independent control of transform unit dimensions to better match the two-dimensional nature of video content and improve encoding flexibility
Data Source
AI summary
This disclosure describes techniques for determining transform partitions in video encoding processes that allow for non-square transform partitions in intra-coded blocks. According to one example of the disclosure, a video coding method comprise partitioning a coding unit into multiple prediction units, and determining a transform partition for each of the prediction units, wherein at least one transform partition is a non-square partition.