Decoder-Side Intra Mode Derivation for Non-Square Video Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like HEVC and VVC face challenges in efficiently deriving intra prediction modes for video blocks, particularly for non-square blocks with varying aspect ratios, leading to suboptimal compression and increased bit rates.
Innovation Solution
The proposed method involves decoder-side intra mode derivation (DIMD) using Histogram of Gradients (HoG) from reconstructed neighbor samples, combining multiple HoGs to improve prediction accuracy and reduce computational complexity, and employing template-based methods to implicitly determine intra prediction modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional intra prediction modes are used in existing video coding standards, then the coding process is simple, but prediction accuracy is suboptimal for non-square blocks with varying aspect ratios
Solution Approach 1:
The patent applies dynamics by making the intra prediction mode selection adaptive to the current block's characteristics. The decoder derives intra prediction modes dynamically based on the aspect ratio of the current block and the prediction mode of neighboring blocks, rather than using fixed prediction modes. This allows the prediction accuracy to adapt to different block shapes and content characteristics, resolving the contradiction between simple coding and suboptimal prediction accuracy.
Solution Approach 2:
The patent changes parameters by introducing new parameters for intra prediction mode derivation that depend on the aspect ratio of the current block and the mode of neighboring blocks. By changing the prediction mode parameter based on these conditions, the system achieves better prediction accuracy for non-square blocks without requiring complex additional processing, thus resolving the contradiction between prediction accuracy and coding complexity.
2Measurement precision
If more sophisticated intra prediction mode derivation is applied, then prediction accuracy improves, but computational complexity increases
Solution Approach 1:
The patent applies local quality by deriving intra prediction modes specifically for blocks that benefit from it, based on their aspect ratio and neighboring block modes. Rather than applying complex derivation to all blocks uniformly, the method selectively improves prediction accuracy where needed (for non-square blocks with varying aspect ratios) while keeping the coding simple for other cases, thus resolving the contradiction between prediction accuracy and computational complexity.
Solution Approach 2:
The patent uses preliminary action by deriving the intra prediction mode at the decoder side using already available information (reconstructed neighboring blocks and current block dimensions) before actual decoding. This preliminary derivation avoids the need for complex encoder-side processing and transmission of additional mode information, achieving better prediction accuracy without significantly increasing overall computational complexity.
3Loss of substance
If adaptive intra prediction modes are derived at decoder side, then bit rate is reduced, but processing complexity at decoder increases
Solution Approach 1:
The patent applies self-service by enabling the decoder to autonomously derive intra prediction modes using only the information already present in the bitstream (neighboring block reconstruction data and current block dimensions). The decoder serves itself by generating the necessary prediction mode information without requiring additional encoder-side processing or bitstream transmission, thus reducing bit rate while keeping decoder processing complexity manageable.
Solution Approach 2:
The patent uses copying by deriving intra prediction modes from the reconstruction of neighboring blocks, effectively copying the prediction approach from already-decoded adjacent blocks. This copying mechanism allows the decoder to generate accurate prediction modes for the current block using existing decoded data, reducing the need for additional bitrate information while maintaining reasonable processing complexity.
Data Source
AI summary
A video coding method using decoder-side intra mode derivation (DIMD) based on multiple histograms of gradients (HoGs) is provided. A video coder receives data to be encoded or decoded as a current block of a current picture of a video. The video coder derives a plurality of HoGs for different intra prediction angles. Different HoGs are derived based on different sets of reconstructed pixels neighboring the current block. The video coder selects a HoG from the plurality of HoGs. The video coder identifies one or more intra prediction modes based on the selected HoG. The video coder generates an intra-prediction of the current block based on the identified one or more intra prediction modes. The video coder encodes or decodes the current block by using the generated intra-prediction.