Video Coefficient Context Modeling With Distance-Weighted Bin Classes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding methods use a fixed local pattern to calculate context models for entropy coding, which may not achieve the most efficient compression due to inadequate consideration of the distance from the current block and fixed number of context models, leading to inaccurate probability predictions and increased bit usage.
Innovation Solution
Adaptive context modeling that takes into account the distance from the current block, uses cross-bin dependencies, and flexibly controls the local template size and number of context models to improve the accuracy of entropy coding by dynamically selecting the most appropriate context model based on the video characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed local pattern is used to calculate the context model, then the calculation process is simple, but the compression efficiency is insufficient due to inaccurate probability predictions
Solution Approach 1:
The patent transforms the fixed local pattern into a dynamic adaptive pattern that adjusts based on video content characteristics. The context model calculation adapts to different regions and content types, allowing the system to maintain calculation simplicity while significantly improving compression efficiency through content-aware probability predictions.
Solution Approach 2:
The patent changes the parameters of the local pattern by introducing distance-weighted coefficients and adaptive template sizes. Instead of using a fixed pattern, the system dynamically adjusts the pattern parameters based on the distance from the current block and the local video content characteristics, resolving the contradiction between simplicity and efficiency.
2Device complexity
If a fixed number of context models is used, then the model management is simple, but the prediction accuracy is insufficient for diverse video characteristics
Solution Approach 1:
The patent implements a dynamic context model selection mechanism that adapts the number of context models based on video content characteristics. The system automatically adjusts the model count and selection based on local patterns, maintaining simple model management while achieving high prediction accuracy for diverse video content through adaptive model allocation.
Solution Approach 2:
The patent applies different numbers and types of context models to different regions of the video content based on local characteristics. Instead of using a uniform fixed number of models throughout, the system tailors the context model configuration to local video patterns, improving prediction accuracy while keeping overall system complexity manageable through localized adaptation.
3Speed
If distance from the current block is not considered, then the context model calculation is fast, but the probability prediction is inaccurate
Solution Approach 1:
The patent introduces distance as a dynamic parameter in the context model calculation by applying distance-weighted coefficients to neighboring blocks. This allows the system to maintain fast calculation speed through efficient distance-based weighting while significantly improving probability prediction accuracy by giving appropriate emphasis to closer blocks that have higher correlation with the current block.
Data Source
AI summary
In some embodiments, a method determines a plurality of classes of bins that are used to determine a context model for entropy coding of a current block in a video. The method calculates a first value for a first class of bins in the plurality of classes of bins and calculates a second value for a second class of bins in the plurality of classes of bins. The first value for the first class of bins is weighted by a first weight to generate a weighted first value and the second value for the second class of bins is weighted by a second weight to generate a weighted second value. The method then selects a context model based on the first weighted value and the second weighted value.