Unified Significance Map Coding With Shared Contexts Across Transforms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as MPEG-4 AVC and KTA, face challenges in effectively capturing the difference in coefficient distributions across various transform sizes due to limited context models and separate context sharing methods for different transforms, which complicates the extension to future standards with more transforms.
Innovation Solution
A unified rule-based approach is proposed for generating context sharing maps that adapt to transform sizes, allowing multiple transform coefficient positions to share contexts, simplifying the design of encoders and decoders and enabling easier extension to future standards.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate context models are used for different transform sizes, then the coefficient distributions can be captured more accurately, but the device complexity and number of context models increase
Solution Approach 1:
The patent merges context models by having different transform sizes (4x4, 8x8, 16x16, etc.) share common context models. Specifically, transform coefficients are grouped into frequency bands, and context models are shared across transforms of different sizes within the same frequency band, reducing the total number of context models while maintaining effective modeling.
Solution Approach 2:
The patent creates universal context models that can be applied across multiple transform sizes. The same context model serves multiple transform types by adapting to different transform sizes through frequency band grouping, making the context modeling system multi-functional and reducing redundancy.
2Measurement precision
If more context models are created to accommodate different transform sizes, then the coding accuracy improves, but the ease of extension to future standards deteriorates
Solution Approach 1:
The patent establishes a universal framework where context models are not tied to specific transform sizes but are instead applicable across multiple transform types. This universality makes it easy to extend the system to future standards with new transform sizes, as the same context modeling approach can be reused without creating new models for each transform type.
Solution Approach 2:
The patent segments transform coefficients into frequency bands and assigns context models based on frequency characteristics rather than transform size. This segmentation approach allows the system to handle different transform sizes uniformly by grouping coefficients with similar frequency characteristics, facilitating easier extension to future standards.
3Device complexity
If limited context templates are used, then the model cost is reduced, but the measurement precision of conditional probability estimation deteriorates
Solution Approach 1:
The patent applies local quality by selecting context templates based on the local characteristics of transform coefficients, specifically their frequency band and position within the transform block. Rather than using a uniform limited set of templates, the system adapts template selection to local coefficient properties, improving probability estimation accuracy while keeping the overall model cost manageable.
Data Source
AI summary
Methods and apparatus are provided for unified significance map coding. An apparatus includes a video encoder (400) for encoding transform coefficients for at least a portion of a picture. The transform coefficients are obtained using a plurality of transforms. One or more context sharing maps are generated for the transform coefficients based on a unified rule. The one or more context sharing maps are for providing at least one context that is shared among at least some of the transform coefficients obtained from at least two different ones of the plurality of transforms.