Matrix-Based Intra Prediction With Block-Derived References
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding/decoding technologies face challenges in optimizing down-sampling filters for cross-component prediction, particularly when dealing with various color formats and sub-sampling factors, leading to inefficiencies in encoding/decoding processes.
Innovation Solution
The method involves determining intra prediction modes for current blocks, deriving reference sample lengths based on block sizes, and using matrix-based intra prediction (MIP) to generate prediction blocks, with specific handling for chroma blocks through cross-component prediction and down-sampling of neighboring samples, including the use of inner samples when neighboring samples are unavailable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If various types of down-sampling filters are selectively used according to the location of down-sampling and the effectiveness of neighboring pixels, then the accuracy of cross-component prediction is improved, but the complexity of implementation in software and hardware increases
Solution Approach 1:
The patent changes the parameter of down-sampling filter selection from location-dependent to uniform. Instead of selecting different filters based on down-sampling location and neighboring pixel effectiveness, the patent applies a uniform down-sampling process using inner samples regardless of position, simplifying the implementation while maintaining prediction accuracy
Solution Approach 2:
The patent extracts and uses only the inner samples of the luma block for cross-component prediction, ignoring other neighboring pixels. This extraction of a specific subset of pixels simplifies the down-sampling process by focusing only on the most effective samples, reducing implementation complexity while maintaining accuracy
2Adaptability or versatility
If down-sampling is performed with consideration of various color formats and sub-sampling factors, then the adaptability of the encoding/decoding process is improved, but the complexity of the down-sampling process increases
Solution Approach 1:
The patent creates a universal down-sampling process that works across various color formats and sub-sampling factors. By using inner samples from the luma block regardless of the specific color format (4:4:4, 4:2:2, 4:2:0, etc.), the patent achieves multi-format adaptability with a single simplified down-sampling approach, reducing process complexity while maintaining versatility
3Measurement precision
If neighboring samples are used for cross-component prediction, then the prediction accuracy is improved, but the problem arises when neighboring samples are unavailable at block boundaries
Solution Approach 1:
The patent uses inner samples of the luma block as an intermediary source when traditional neighboring samples are unavailable. Instead of relying on external neighboring blocks that may not exist at boundaries, the patent mediates the prediction process by using samples from within the same block, ensuring reliability while maintaining accuracy
Solution Approach 2:
The patent enables the current block to serve itself for prediction by using its own inner samples. When neighboring samples are unavailable, the block performs self-prediction using its internal luma samples, eliminating dependency on external references and ensuring reliable prediction regardless of block position
Data Source
AI summary
Disclosed herein is an image encoding/decoding method. A method of decoding an image includes determining an intra prediction mode of a current block, and generating a prediction block of the current block, by performing prediction based on the intra prediction mode. When the intra prediction mode of the current block is a matrix-based intra prediction (MIP) mode, a length of a reference sample used for intra prediction is derived based on a length of the current block.