Context Reduction in Last Significant Coefficient Position Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression standards, such as H.264 and HEVC, face inefficiencies in reducing the number of contexts used when coding last transform positions, which affects coding efficiency and quality.
Innovation Solution
The proposed solution involves encoding the position of the last non-zero coefficient within a video block as a string of binary values, with context index values derived from bin index values using a lookup table, allowing context index values to be shared across blocks of different widths, and employing entropy encoding techniques like CABAC for efficient coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the number of contexts is reduced in last significant coefficient position coding, then device complexity is reduced, but coding efficiency deteriorates
Solution Approach 1:
The patent merges context models across different block widths by deriving context indices from normalized bin indices. Instead of maintaining separate context models for each block width (4x4, 8x8, 16x16, 32x32), the invention normalizes the bin index by dividing by the block width and uses this normalized value to select from a shared set of context models. This merging reduces the total number of contexts while preserving coding efficiency through the normalization approach.
Solution Approach 2:
The patent creates universal context models that can be applied across multiple block widths. By deriving context indices from normalized bin indices rather than using width-specific contexts, a single context model serves multiple functions across different transform unit sizes. This universal approach reduces device complexity by eliminating redundant width-specific context models while maintaining adaptability through the normalization mechanism.
2Device complexity
If context index values are shared across blocks of different widths, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent changes the parameter used for context selection from the raw bin index (which varies with block width) to a normalized bin index (bin index divided by block width). This parameter transformation allows the same context index to be appropriately applied across different block widths. The normalization preserves the relative position information needed for precise coding while enabling context sharing, thus maintaining measurement precision despite using shared context models.
Data Source
AI summary
A method of video encoding includes encoding a position of a last non-zero coefficient within a video block having a first width. The position of the last non-zero coefficient is provided as a string of binary values, wherein each binary value in said string corresponds to a bin index value. The method also includes determining a context index value for each bin index value, wherein the context index value for each bin index value is configured to be shared across two or more blocks of different widths.