MIP-Based LFNST Selection for Intra Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The LFNST technology in H.266/VVC lacks variability in transform selection for non-traditional intra prediction modes, particularly Matrix-based Intra Prediction (MIP), leading to reduced encoding efficiency.
Innovation Solution
Introduce a Matrix-based Intra Prediction (MIP) parameter to determine an intra prediction value and select a Low-Frequency Non-Separable Transform (LFNST) kernel flexibly, signaling the LFNST index in the video bitstream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If LFNST technology uses fixed transform set selection based on traditional intra prediction mode, then implementation is simple, but encoding efficiency is reduced for non-traditional modes like MIP
Solution Approach 1:
The patent changes the parameter used for LFNST transform set selection from traditional intra prediction mode to MIP parameter (mipModeId). This allows the transform selection to adapt to non-traditional intra prediction modes like MIP, improving encoding efficiency by selecting more appropriate transform sets for different prediction modes while maintaining a relatively simple implementation through parameter-based selection rules.
2Adaptability or versatility
If LFNST transform set is fixed for non-traditional intra prediction modes, then implementation is straightforward, but applicability and flexibility are limited
Solution Approach 1:
The patent creates a universal transform selection mechanism that works for both traditional and non-traditional intra prediction modes. By using MIP parameter as the basis for transform set selection and providing default transform sets for modes without explicit MIP parameters, the system achieves broad applicability across different prediction modes while maintaining a unified, relatively simple selection framework.
Data Source
Figure 1~2A
Figure 2B~3
Figure 4~6A
AI summary
Disclosed are a transformation method, an encoder, a decoder, and a storage medium. The transformation method applied to an encoder includes: determining a prediction mode parameter of a current block; determining a Matrix-based Intra Prediction (MIP) parameter when the prediction mode parameter indicates that MIP is used for the current block to determine an intra prediction value; determining an intra prediction value of the current block according to the MIP parameter and calculating a residual value between the current block and the intra prediction value; determining a Low-Frequency Non-Separable Transform (LFNST) transform kernel used for the current block according to the MIP parameter when an LFNST is used for the current block, setting an LFNST index, and signalling the LFNST index in a video bitstream; and performing a transform processing on the residual value by using the LFNST transform kernel.