Video Compression Parity Hiding Reduces Context-Coded Bins

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding standards, such as H.264/AVC and developing HEVC, face inefficiencies in encoding and decoding transform coefficients, particularly in the high percentage of bitstream data occupied by quantized transform coefficients, which can be improved by adaptive threshold settings and sign data hiding techniques.

Innovation Solution

The proposed method involves using parity to hide information about the upper left coefficient in a coefficient group, conditional decoding of significant-coefficient and greater-than-one flags based on the parity of the upper left coefficient, and reducing the maximum number of greater-than-one flags to enhance encoding and decoding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If quantized transform coefficients are encoded using conventional context modeling, then compression quality is maintained, but the number of context-coded bins increases significantly (occupying 30-80% of bitstream data)

Engineering Contradiction:
Improvecompression efficiencyVSAvoidnumber of context-coded bins
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent extracts and removes redundant context-coded bins from the encoding process by identifying that certain coefficient positions (particularly in the upper-left region of transform blocks) do not require separate context modeling. This is achieved by applying adaptive thresholds that eliminate the need for context-coded bins in regions where coefficients are likely to be zero or follow predictable patterns, thereby reducing the overall number of bins without sacrificing compression quality.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different encoding strategies to different regions of the transform block. Adaptive thresholds are applied specifically to the upper-left coefficient groups where redundancy is highest, while other regions maintain conventional encoding. This localized approach reduces the number of context-coded bins in regions where it matters most while preserving compression efficiency in regions where detailed encoding is still necessary.

Inventive Principle:
Principle #3Local quality

2Productivity

If adaptive threshold settings are applied to reduce context-coded bins, then compression efficiency improves, but encoding complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidencoding complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent performs preliminary analysis of the transform coefficient patterns before the main encoding process. By pre-determining which coefficient regions will benefit from adaptive thresholding based on statistical properties and prediction models, the system prepares encoding parameters in advance. This preliminary action allows the main encoding loop to simply apply pre-computed thresholds rather than performing complex real-time analysis, thereby improving compression efficiency without proportionally increasing encoding complexity.

Inventive Principle:
Principle #10Preliminary action

3Device complexity

If the number of greater-than-one flags is reduced, then the number of context-coded bins decreases, but measurement precision of coefficient magnitudes may be affected

Engineering Contradiction:
Improvenumber of context-coded binsVSAvoidcoefficient magnitude precision
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies greater-than-one flags selectively rather than universally. By using adaptive thresholds, the system applies these flags only in regions where coefficient magnitudes are likely to exceed the threshold, rather than encoding them for all coefficients. This partial application reduces the number of context-coded bins while maintaining precision where it is actually needed, based on statistical analysis of coefficient distributions in different transform block regions.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2680580B1Reduced worst-case context-coded bins in video compression with parity hiding
Publication Date: 2016.06.01 BLACKBERRY LTD
  • EP2680580B1 patent drawingFigure 1
  • EP2680580B1 patent drawingFigure 2
  • EP2680580B1 patent drawingFigure 3

AI summary

Methods and devices for reconstructing coefficient levels from a bitstream of encoded video data for a coefficient group in a transform unit. Parity hiding is used to signal the parity of the upper left coefficient of the coefficient group. Based on the parity, either the significant coefficient flag or the greater-than-one flag is encoded/decoded for that coefficient, but not both. The greater-than-one flag is encoded/decoded irrespective of whether a maximum number of greater-than-one flags have been encoded/decoded with respect to the other coefficients in the coefficient group.