Transform Coefficient Coding With Dependent Quantization Passes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing coding technologies face challenges in achieving efficient coding of transform coefficients due to increased coding complexity and reduced information reservoir when using dependent quantization combined with context adaptive entropy coding, which affects the interrelationship between data items and context derivation.
Innovation Solution
The proposed solution involves decoding and encoding transform coefficients using context-adaptive binary arithmetic decoding/encoding in a sequence of passes, where flags and remainder values are used to bi-split the value domain of quantization indexes, and a state transitioning mechanism is employed to determine reconstruction levels based on previous coefficients, allowing for efficient coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If dependent quantization is used to improve coding efficiency, then coding efficiency is improved, but coding complexity increases
Solution Approach 1:
The patent segments the quantization process into multiple passes (first pass, second pass, third pass) where different types of syntax elements are coded in each pass. This segmentation allows the decoder to progressively refine the reconstruction of transform coefficients without requiring all dependency information to be available simultaneously, thereby reducing coding complexity while maintaining coding efficiency.
Solution Approach 2:
The patent performs preliminary quantization in the first pass to generate initial syntax elements, then uses these results in subsequent passes. The first pass codes absolute values of quantization indexes, and subsequent passes code differences or refinements. This preliminary action allows later passes to operate with reduced complexity since they only need to code deviations from already-determined values.
2Productivity
If dependent quantization is used to improve coding efficiency, then coding efficiency is improved, but information reservoir for context modeling is reduced
Solution Approach 1:
The patent transitions from coding all quantization indexes simultaneously (one-dimensional approach) to coding them across multiple passes (adding the time/dimension of passes). This dimensional change allows context modeling to operate on a richer set of information available at each pass stage, rather than being constrained by the need to model all dependencies in a single pass.
Solution Approach 2:
By performing preliminary coding of certain syntax elements in early passes, the patent creates a foundation of known information that can be used for context modeling in later passes. This preliminary action ensures that context models have access to sufficient information reservoir while still achieving efficient coding through the progressive refinement approach.
3Manufacturing precision
If finer quantization is used to decrease distortion, then distortion is decreased, but bitrate increases
Solution Approach 1:
The patent applies different coding strategies to different types of syntax elements and different passes. In the first pass, absolute values are coded with higher precision, while in subsequent passes, only the necessary differences or refinements are coded with appropriate precision. This local quality approach ensures that quantization precision is maintained where needed while minimizing the bitrate required for transmitting the quantization parameters.
Data Source
AI summary
Concepts are presented which achieve a more efficient coding of coefficients of a transform block by use of dependent quantization and context adaptive entropy coding or achieve a coding of coefficients of a transform block in a manner which allows a more efficient coding even if a usage of dependent quantization is combined with the usage of context adaptive entropy coding.