Unified Tree Coding for Luminance-Chrominance Block Flags
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In image coding and decoding methods, there is a demand to reduce the data amount stored in memory for temporarily holding data used in coding or decoding, particularly due to the increased data requirements for managing transform sizes and coefficients.
Innovation Solution
The method involves splitting an input image signal into coding units, performing orthogonal transform and quantization on prediction error signals, and coding management information and quantization coefficients into a tree structure, where each transform unit corresponds to a leaf node, allowing for continuous coding of management information and quantization coefficients, reducing the need to store information from other transform units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If management information and transform coefficients are coded into separate tree structures, then the flexibility in selecting transform sizes is improved, but the data amount in memory for temporarily holding data increases
Solution Approach 1:
The patent combines the coding of management information (CBFs) and transform coefficients into a single unified tree structure. By merging these two previously separate coding processes, the patent eliminates the need to store management information of other transform units in memory, thereby reducing memory data amount while preserving the flexibility to select different transform sizes through the tree structure's inherent adaptability
2Manufacturing precision
If management information of all transform units is stored in memory, then the coding accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent extracts only the necessary transform coefficients from the coded stream and processes them immediately without storing all management information in memory. By taking out only the essential data needed for accurate coding and processing it on-the-fly, the patent maintains coding accuracy while significantly reducing memory storage requirements and device complexity
3Reliability
If separate tree structures are used for management information and transform coefficients, then the reliability of decoding is improved, but the loss of time in processing increases
Solution Approach 1:
The patent implements continuous coding of management information and transform coefficients in a single unified tree traversal process. This continuous action eliminates the need for separate processing stages and intermediate storage operations, thereby maintaining decoding reliability through complete data processing while reducing overall processing time by eliminating redundant operations
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image coding method, comprising: subtracting a prediction signal (181) from the input image signal (191) for each coding unit (CU), thereby generating respective prediction error signals (111); performing orthogonal transform and quantization on a corresponding one of the prediction error signals (111) for each transform unit (TU), eventually generating quantization coefficients (131); and coding pieces of management information indicating a structure of the transform units (TU) and the quantization coefficients (131) into a tree structure. Each of the transform units corresponds to a corresponding one of leaf nodes in the tree structure. In the coding, for each leaf node, management information and a quantization coefficient (131) are coded, eventually generating a coded signal (191) in which the coded management information and the coded quantization coefficient are arranged in succession for each leaf node.