Intra Prediction Mode Signaling Complexity Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as HEVC and VVC, face challenges in efficiently signaling intra prediction modes, particularly when the number of modes exceeds a certain limit, leading to increased complexity and bitstream size.
Innovation Solution
The proposed method involves removing one or two directional modes from the 65 directional modes when the total number of intra prediction modes exceeds a certain allowed number, allowing for fixed-length coding of the remaining modes and reducing the bit-length required for signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of intra prediction modes is increased to capture arbitrary edge directions, then the prediction accuracy is improved, but the complexity of signaling and bitstream size increase
Solution Approach 1:
The 65 directional modes are segmented into two groups: conventional intra prediction modes (within 45 degrees of horizontal and vertical) and wide angular intra prediction modes (beyond 45 degrees). This segmentation allows the system to handle different directional ranges separately, reducing the complexity of signaling by organizing modes into manageable groups with distinct coding strategies.
Solution Approach 2:
The patent introduces a new dimension to mode signaling by adding a wide-angle flag (1-bit) that indicates whether to use a conventional mode or its corresponding wide-angular flipped mode. This binary dimension allows the system to expand from 35 conventional modes to 65 directional modes without proportionally increasing signaling complexity, as the wide-angle information is encoded efficiently using this additional binary dimension.
2Measurement precision
If the number of intra prediction modes is increased to capture arbitrary edge directions, then the prediction accuracy is improved, but the bitstream size increases
Solution Approach 1:
By segmenting modes into conventional and wide-angular groups, the patent enables more efficient entropy coding. The wide-angle flag serves as a coarse-grained indicator that allows subsequent refinement with fewer bits, reducing the overall bitstream size compared to directly encoding 65 modes without segmentation.
Solution Approach 2:
The addition of the wide-angle flag as a new signaling dimension enables compact representation of 65 modes. Instead of requiring log2(65) ≈ 6.37 bits for direct mode indexing, the segmented approach with wide-angle flag reduces the average bit requirement by exploiting the hierarchical structure, thereby decreasing bitstream size while maintaining full 65-mode functionality.
3Adaptability or versatility
If wide angular intra prediction modes are added beyond 45 degrees, then the directional coverage is improved, but the number of signaling parameters increases
Solution Approach 1:
The patent uses the concept of flipped modes where wide-angular modes are derived by flipping conventional modes across the 45-degree diagonal. Instead of defining entirely new mode directions, the system inverts the conventional mode indexing and applies a flip transformation, reducing the number of independent parameters that need to be signaled while achieving extended directional coverage.
Solution Approach 2:
The wide-angle flag introduces a binary dimension that toggles between conventional and wide-angular interpretations of mode indices. This additional dimension allows the system to double the effective directional coverage without doubling the base mode set, thereby improving adaptability while controlling the growth of signaling parameters through efficient use of the new binary dimension.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
A method of controlling intra prediction for decoding or encoding of a video sequence, is performed by at least one processor and includes obtaining intra prediction modes including directional modes respectively corresponding to angular prediction directions, a first amount of one or more of the directional modes being excluded from the intra prediction modes based on a second amount of the intra prediction modes and a third amount of most probable modes (MPMs). The method further includes selecting, as the MPMs, two or more of the intra prediction modes from which the one or more of the directional modes are excluded, and selecting, for decoding the video sequence, one of the intra prediction modes from which the one or more of the directional modes are excluded.