Adaptive Multitree Subdivision for Two-Dimensional Signal Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video coding standards are limited in their ability to sub-divide pictures into blocks, leading to inefficient use of bit rate for signaling prediction parameters due to the mismatch between block boundaries and object boundaries, especially for arbitrarily shaped objects, which increases encoding complexity and reduces coding efficiency.
Innovation Solution
A coding scheme that spatially divides an array of information samples into tree root regions and recursively subdivides them into smaller simply connected regions using multi-tree subdivision, allowing for a better compromise between encoding complexity and rate-distortion performance by including maximum region size and multi-tree subdivision information in the data stream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If pictures are decomposed into blocks with fixed or signaled subdivision, then coding parameters can be associated with blocks for prediction, but the block boundaries do not match object boundaries for arbitrarily shaped objects, increasing encoding complexity and reducing coding efficiency
Solution Approach 1:
The patent applies segmentation by dividing the picture into multiple levels of blocks (macroblocks, sub-macroblocks, and further subdivisions) with different granularities. Each level serves specific coding purposes, allowing flexible adaptation to object shapes through multi-scale partitioning rather than uniform blocking.
Solution Approach 2:
The patent introduces dynamic subdivision where blocks can be further split into smaller regions based on content characteristics. The subdivision is not fixed but adapts dynamically to the actual object boundaries and spatial distribution, enabling better alignment with arbitrarily shaped objects while managing complexity through selective refinement.
2Adaptability or versatility
If more syntax elements are transmitted to signal subdivision for blocks, then better adaptation to object boundaries is achieved, but the bit rate for signaling increases
Solution Approach 1:
The patent applies local quality by allowing different regions of the picture to have different subdivision granularities. Areas with complex object boundaries receive finer subdivision with more syntax elements, while uniform areas use coarser blocking with fewer syntax elements, optimizing the balance between adaptation and bit rate overhead.
Solution Approach 2:
The patent implements partial subdivision where not all blocks are subdivided to the maximum level. Instead, subdivision is applied selectively to regions where it provides actual benefit, avoiding the excessive signaling overhead that would result from uniform fine-grained subdivision across the entire picture.
3Measurement precision
If blocks are subdivided into smaller blocks for prediction, then prediction accuracy improves, but the amount of side information for signaling subdivision increases
Solution Approach 1:
The patent resolves the contradiction by adding a hierarchical dimension to the block structure. Instead of a single level of fine blocks, it creates multiple hierarchical levels (macroblocks → sub-macroblocks → smaller regions) where prediction can be performed at appropriate granularities, reducing the total side information required compared to a flat fine-grained structure.
Solution Approach 2:
The patent applies the nested doll principle by nesting smaller blocks within larger blocks in a hierarchical structure. Each level of nesting provides prediction at its scale, and the nested structure allows shared syntax elements between parent and child blocks, reducing the total side information rate compared to independent fine blocks.
Data Source
AI summary
Coding schemes for coding a spatially sampled information signal using sub-division and coding schemes for coding a sub-division or a multitree structure are described, wherein representative embodiments relate to picture and/or video coding applications.