Intra Prediction Mode Coding Using Context Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing HEVC standard's intra prediction mode coding is inefficient due to limited prediction angles and complexity, and the current three Most Probable Modes (MPMs) do not accurately correspond to the actual probability distribution of all available intra prediction modes, leading to suboptimal compression efficiency.
Innovation Solution
The proposed solution involves deriving more than three MPMs, using context modeling for decoding MPM indices, and applying an N-tap intra interpolation filter with N greater than 2 to neighboring reconstructed samples, allowing for multiple intra prediction directions per block without additional signaling, and using 1/32-pel accuracy for interpolation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If more than three MPMs are derived and context modeling is used for decoding MPM indices, then compression efficiency is improved and bitstream size is reduced, but encoder and decoder complexity increases
Solution Approach 1:
The patent changes the parameter of MPM count from the conventional three to more than three (e.g., four or five MPMs), and introduces context modeling for decoding MPM indices. This parameter change improves compression efficiency by better matching the actual probability distribution of intra prediction modes, while the complexity increase is managed through systematic implementation of context adapters and probability updates.
Solution Approach 2:
The patent introduces dynamic context modeling where context adapters and probability values are updated based on previously decoded MPM indices and neighboring block information. This dynamic adaptation allows the decoder to efficiently handle the increased MPM count without linearly increasing complexity, as the context model learns and adapts to the local characteristics of the video content.
2Measurement precision
If an N-tap intra interpolation filter with N greater than 2 is applied to neighboring reconstructed samples, then interpolation accuracy is improved and edge direction distribution is better captured, but computational complexity increases
Solution Approach 1:
The patent changes the filter tap count parameter from the conventional 2-tap filter to an N-tap filter where N > 2 (e.g., 3-tap, 4-tap, or 5-tap filters). This parameter change improves interpolation accuracy by better capturing edge direction distributions in the video content. The increased computational complexity is offset by the fact that the filter coefficients are predetermined and can be efficiently implemented using lookup tables or optimized arithmetic operations.
3Reliability
If multiple intra prediction directions per block are allowed without additional signaling, then prediction accuracy is improved, but the decoding process becomes more complex
Solution Approach 1:
The patent makes the MPM list and context modeling mechanism universal and adaptable to multiple intra prediction directions within the same block. By deriving more than three MPMs and using context modeling, the system can handle multiple prediction directions without requiring separate signaling for each direction, thus improving prediction accuracy while keeping the decoding process manageable through unified treatment of multiple modes.
4Device complexity
If the existing HEVC standard's three MPMs are used, then encoder complexity is maintained, but compression efficiency is suboptimal due to inaccurate probability distribution modeling
Solution Approach 1:
The patent changes the MPM count parameter from three to more than three (four or five MPMs), which better represents the actual probability distribution of intra prediction modes. This parameter change improves compression efficiency by reducing the entropy of mode signaling, while encoder complexity increases are managed through efficient derivation methods that leverage neighboring block information and previously decoded modes.
Data Source
AI summary
A device for decoding video data includes a memory configured to store the video data; and one or more processors configured to decode syntax information that indicates a selected intra prediction mode for the block of video data from among a plurality of intra prediction modes. The one or more processors apply an N-tap intra interpolation filter to neighboring reconstructed samples of the block of video data according to the selected intra prediction mode, wherein N is greater than 2. The one or more processors reconstruct the block of video data based on the filtered neighboring reconstructed samples according to the selected intra prediction mode.