Decoder Side Intra Mode Derivation Complexity Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing decoder-side intra-mode determination (DIMD) tool for video coding is computationally complex and increases signaling overhead, especially at coding tree block boundaries, due to the need to test and signal multiple intra-modes, which leads to increased memory access and processing time.
Innovation Solution
The proposed solution involves an apparatus and method for intra-predicting coding blocks by deactivating part or all of the template matching at coding tree block boundaries, reducing computational complexity by determining the intra-prediction mode based on adjacent blocks only when their borders align with coding tree block borders, and omitting signaling for unavailable blocks, thereby reducing memory access and signaling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If decoder-side intra-mode determination (DIMD) tool is used to reduce signaling overhead, then signaling overhead is reduced, but computational complexity increases
Solution Approach 1:
The patent applies local quality by differentiating the treatment of coding blocks based on their spatial location. Specifically, template matching is performed only when the current coding block is not adjacent to a coding tree block boundary, while skipping template matching when adjacent to boundaries. This localized adaptation reduces computational complexity in specific regions without compromising overall prediction accuracy.
Solution Approach 2:
The patent segments the image into coding blocks and identifies which blocks require full template matching based on their relationship with coding tree block boundaries. By dividing the processing task into segments (blocks adjacent to boundaries vs. blocks not adjacent to boundaries), the system can apply different processing intensities to different segments, thereby reducing overall computational complexity.
2Measurement precision
If template matching is performed for all coding blocks, then intra-mode determination accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by performing template matching only for a subset of coding blocks (those not adjacent to coding tree block boundaries) while skipping it for blocks adjacent to boundaries. This partial application reduces processing time while maintaining sufficient accuracy for the majority of blocks, accepting that boundary blocks will use alternative prediction methods.
3Measurement precision
If multiple intra-modes are tested and signaled, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies local quality by adapting the prediction method based on the local position of coding blocks. For blocks adjacent to coding tree block boundaries, the system uses a simplified prediction approach without template matching, while maintaining full template matching and multiple intra-mode testing for blocks not adjacent to boundaries. This localized differentiation maintains prediction accuracy where needed while reducing complexity where appropriate.
Data Source
AI summary
The present disclosure relates to image encoding/decoding applying decoder side intra-prediction mode determination (DIMD). In order to reduce the computational complexity, DIMD for a current coding block based on a template of a neighboring reconstructed coding block is not performed if the border between the current and the reconstructed coding block concurs with a border between coding tree blocks.