Intra Template Matching Decoding With Multi-Block Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital video compression technologies face challenges in improving compression efficiency, particularly in scenarios with complex pixel changes and noise signals, where traditional IntraTMP modes may reduce decoding performance.
Innovation Solution
A decoding method that predicts multiple matching blocks using the IntraTMP mode, allowing for flexible selection of target prediction blocks, and optionally fusing these with other prediction modes to enhance decoding performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional IntraTMP mode is used for prediction, then decoding simplicity is maintained, but prediction accuracy deteriorates in scenarios with complex pixel changes and noise signals
Solution Approach 1:
The patent implements dynamic prediction block selection by determining multiple candidate prediction blocks from different templates (intra templates and inter templates) and dynamically selecting the optimal one based on matching accuracy. This allows the system to adapt to different content characteristics (complex pixel changes vs. simple patterns) while maintaining decoding flexibility and improving prediction accuracy without excessive complexity
Solution Approach 2:
The patent changes the prediction parameter by using multiple templates with different characteristics (intra templates for spatial correlation, inter templates for temporal correlation) and selecting appropriate templates based on content type. This parameter variation enables the system to handle diverse video content effectively, improving prediction accuracy for complex scenarios while maintaining simplicity for standard cases
2Productivity
If multiple matching blocks are predicted and fused, then coding performance is improved, but computational overhead increases
Solution Approach 1:
The patent applies partial action by determining multiple prediction blocks but not necessarily using all of them equally. Instead, it selects the most suitable prediction block(s) based on matching criteria, performing fusion only when beneficial. This approach improves coding performance by utilizing multiple candidates while controlling computational overhead through selective rather than exhaustive processing
Solution Approach 2:
The patent incorporates feedback mechanisms by evaluating the matching accuracy of different prediction blocks and using this information to select or fuse the most appropriate ones. The system feedbacks on prediction quality and adjusts the selection/fusion strategy accordingly, improving coding performance while avoiding unnecessary computational overhead from processing all possible candidates equally
Data Source
Figure 1~2
Figure 3~4
Figure 5(a)~7
AI summary
Provided in the embodiments of the present application are a decoding method, a coding method, decoders, and coders. The decoding method comprises: on the basis of a code stream, determining a residual block of a current block in a current sequence; on the basis of an intra template matching prediction (Intra TMP) mode, predicting at least one matching block of the current block; on the basis of the at least one matching block, determining a target prediction block of the current block; and on the basis of the residual block of the current block and the target prediction block, obtaining a reconstructed block of the current block. In the embodiments of the present application, a decoder can determine the target prediction block on the basis of a single matching block selected by the IntraTMP mode, and can also determine the target prediction block on the basis of a plurality of matching blocks selected by the IntraTMP mode, which not only improves the prediction flexibility of the current block, but also facilitates the decoder in selecting a suitable number of matching blocks to determine the target prediction block, thereby improving the coding and decoding performance.