Intra Prediction Data Derivation for Arbitrary Rectangular Video Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as ITU-T H.265, face limitations in efficiently encoding and decoding video data using arbitrary rectangular video blocks, particularly in terms of quadtree binary tree partitioning and transform coefficient scaling, which can lead to suboptimal compression and decoding performance.
Innovation Solution
The method involves determining an intra prediction mode for a current block as the planar prediction mode and deriving intra prediction data based on the block's dimensions and adjacent reconstructed samples, allowing for efficient partitioning and encoding of video data using arbitrary rectangular blocks, and applying adaptive scaling of transform coefficients based on block dimensions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If quadtree binary tree partitioning is used in current video coding standards, then video data can be encoded and decoded, but compression and decoding performance becomes suboptimal due to limitations in handling arbitrary rectangular video blocks
Solution Approach 1:
The patent applies segmentation by dividing the video picture into arbitrary rectangular blocks rather than using fixed quadtree binary tree partitioning. This allows the video data to be segmented into blocks of varying sizes and shapes, enabling better adaptation to different video content characteristics and improving compression efficiency through more flexible region-of-interest encoding.
Solution Approach 2:
The patent implements dynamic block partitioning where the size and shape of coding blocks can vary adaptively across different regions of the video picture. This dynamic approach allows the encoder to adjust block dimensions based on local content characteristics, such as edges, textures, or motion patterns, thereby optimizing compression performance for each specific region.
2Measurement precision
If transform coefficient scaling is applied in current video coding standards, then transform coefficients can be processed, but decoding performance becomes suboptimal due to fixed scaling approaches
Solution Approach 1:
The patent applies local quality by implementing position-dependent scaling for transform coefficients, where different scaling factors are applied to coefficients at different positions within the block. This local differentiation allows for more precise reconstruction of video data by adapting the scaling behavior to local frequency characteristics, thereby improving decoding precision without requiring overly complex global scaling mechanisms.
3Productivity
If fixed video block structures are used in existing standards, then encoding and decoding is simplified, but compression efficiency is reduced due to inability to handle arbitrary rectangular blocks
Solution Approach 1:
The patent implements dynamic block partitioning where the size and shape of coding blocks can vary adaptively across different regions of the video picture. This dynamic approach allows the encoder to adjust block dimensions based on local content characteristics, such as edges, textures, or motion patterns, thereby optimizing compression performance for each specific region.
Solution Approach 2:
The patent changes the parameters of video blocks from fixed dimensions to variable dimensions, allowing arbitrary rectangular blocks with different widths and heights. This parameter change enables the encoding system to adapt block sizes to match the actual content structure, improving compression efficiency by reducing redundancy in homogeneous regions and preserving detail in complex regions.
Data Source
AI summary
A method of deriving intra prediction data based on a planar prediction mode includes the following operations: obtaining multiple first reconstructed row samples and a second reconstructed row sample in an adjacent above row to a current block, the first reconstructed row samples extending from first to last columns of the current block, the second reconstructed row sample corresponding to an adjacent right column to the current block; obtaining multiple first reconstructed column samples and a second reconstructed column sample in an adjacent left column to the current block, the first reconstructed column samples extending from first to last rows of the current block, the second reconstructed column sample corresponding to an adjacent below row to the current block; and deriving the intra prediction data corresponding to the current block based on a width and a height of the current block and the obtained reconstructed row and column samples.