Image Encoding Dependency Tree for Anti-Causal Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression techniques, such as H.264, are limited in prediction capabilities as they only utilize blocks in the causal zone for prediction, ignoring potentially similar blocks in the anti-causal zone due to differences in writing and reading order, which can lead to suboptimal compression efficiency.
Innovation Solution
A method that constructs a dependency tree allowing for prediction using both causal and anti-causal zone blocks, enabling the use of neighboring blocks not yet considered in the writing order, and encodes this information for flexible decoding paths, separate from the reading path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If only causal zone blocks are used for prediction, then decoding order is simplified and follows writing path order, but prediction efficiency is reduced due to unavailable anti-causal zone blocks
Solution Approach 1:
The patent segments the prediction process into two independent parts: encoding phase and decoding phase. During encoding, multiple reference blocks including anti-causal zone blocks are identified and stored with their spatial coordinates. During decoding, the dependency tree is traversed to determine the correct decoding order, allowing reference blocks to be accessed in any spatial order while maintaining proper dependency relationships. This segmentation enables efficient prediction without increasing decoding complexity.
Solution Approach 2:
The patent performs preliminary action by pre-identifying and storing all reference blocks (including anti-causal zone blocks) during the encoding phase, along with their spatial coordinates and dependency relationships. This preliminary identification allows the decoder to efficiently retrieve reference blocks in the correct order during decoding, without needing to perform complex real-time analysis.
2Adaptability or versatility
If writing order and reading order are kept identical, then implementation is simpler, but prediction possibilities are limited to causal zone blocks
Solution Approach 1:
The patent introduces a new dimension by separating the writing path order from the reading/decoding path order. The encoding process writes blocks in one order while storing dependency information, and the decoding process reads blocks in a different order determined by traversing the dependency tree. This dimensional separation allows anti-causal zone blocks to be used as references while maintaining implementation feasibility through structured dependency management.
Solution Approach 2:
The patent introduces a dependency tree as an intermediary structure that mediates between the writing order and reading order. The dependency tree stores the relationships between blocks and references, allowing the decoder to determine the correct reading order without directly following the writing path. This intermediary enables flexible prediction possibilities while managing implementation complexity through structured data organization.
3Loss of substance
If anti-causal zone blocks are used for prediction, then compression efficiency improves, but additional information must be stored and decoded out of writing path order
Solution Approach 1:
The patent extracts only the essential information needed for decoding by storing the spatial coordinates and dependency relationships of reference blocks during encoding. Instead of storing complete reference block data, only the necessary metadata (coordinates and dependency links) is extracted and stored in the dependency tree, minimizing additional information storage requirements while enabling efficient retrieval during decoding.
Data Source
Figure 1~2
Figure 3A~3B
Figure 4~5
AI summary
The invention relates to a technique for encoding image data, and to a corresponding decoding technique. The invention comprises implementing a prediction of data of a current block (2) based on at least one already encoded block (1, 6) or reference block, wherein said prediction step takes into account a set of encoded blocks adjacent to said current block and including blocks that have not yet been considered according to said writing path order. The encoding includes the steps of: generating a dependence tree associating a current block with at least one reference block from which the predicted block associated with the current block has been predetermined; and inserting information representative of said dependence tree into said set of data associated with each block.