Matrix-Based Intra Prediction Offsets for Variable Block Sizes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing Matrix-based Intra Prediction (MIP) mode in video encoding requires significant storage space and increases time complexity due to varying parameters for luma blocks of different sizes, affecting encoding and decoding efficiency.
Innovation Solution
A method where the first offset is determined based on the size of the current block, using a correspondence relation between pre-stored indexes and offsets, reducing the need for complex parameter storage and invocation, and enabling efficient encoding and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parameters are stored for luma blocks of different sizes in MIP mode, then prediction accuracy is improved, but storage space requirement increases
Solution Approach 1:
The patent changes the parameter representation by deriving the offset dynamically from the block size and a constant first offset, rather than storing separate parameters for each block size. This transforms the storage requirement from multiple size-specific parameters to a single constant offset value that adapts to different block sizes through calculation.
2Measurement precision
If parameters are stored for luma blocks of different sizes in MIP mode, then prediction accuracy is improved, but encoding and decoding time increases
Solution Approach 1:
The patent performs preliminary action by establishing a constant first offset that can be combined with block size information to derive the second offset on-the-fly. This pre-established constant eliminates the need for complex parameter lookup or calculation during encoding and decoding, reducing processing time while maintaining prediction accuracy.
3Adaptability or versatility
If a large quantity of parameters are stored for different block sizes, then MIP mode prediction capability is improved, but device complexity increases
Solution Approach 1:
The patent applies universality by creating a single constant first offset that serves all block sizes. This universal parameter can be combined with any block size to generate the appropriate second offset, eliminating the need for separate parameters for different block sizes and simplifying the overall system structure.
Data Source
AI summary
Provided are a picture encoding and decoding method, and a storage medium. The picture encoding method includes: determining a size of a current block; when encoding the current block by using a Matrix-based Intra Prediction (MIP) mode, calculating a second offset based on the size of the current block, a first offset and reconstructed values of adjacent pixels corresponding to the current block, wherein the second offset is a value used to perform prediction through the MIP mode and the first offset is a constant; calculating a second prediction value of a pixel at a preset position in the current block by using the second offset, wherein the preset position is a particular position in the current block; performing filtering on the second prediction value to obtain first prediction values of all pixels in the current block.