Transform Coefficient Level Coding for Pipelined Video Decoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding standards, such as H.264/AVC and the developing HEVC, face inefficiencies in encoding and decoding transform coefficients due to throughput issues at the upper-left and bottom-right corners of coefficient groups, particularly with diagonal scan orders, leading to pipelining problems and suboptimal compression efficiency.

Innovation Solution

The use of distinct context templates specific to horizontal and vertical scan orders ensures that neighboring positions are at least three scan order positions previous to the current position, allowing for pipelining by fully processing required data in the context template, and adaptive threshold-based level decoding is employed to optimize coefficient level encoding and decoding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If diagonal scan order is used for encoding transform coefficients, then compression efficiency is improved, but throughput and pipelining performance deteriorate due to bottlenecks at upper-left and bottom-right corners

Engineering Contradiction:
Improvecompression efficiencyVSAvoidthroughput
Core Design Contradiction:
Loss of informationVSProductivity

Solution Approach 1:

The transform coefficient block is divided into multiple sub-blocks, and different scan orders are applied to different sub-blocks. Specifically, diagonal scan order is used for some sub-blocks to maintain compression efficiency, while horizontal or vertical scan orders are used for other sub-blocks to improve throughput and eliminate pipelining bottlenecks at corner positions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different scan orders are applied locally to different regions of the transform coefficient block based on their specific characteristics. The upper-left and bottom-right corner sub-blocks use horizontal or vertical scan orders to avoid pipelining issues, while other regions may benefit from diagonal scan order for better compression.

Inventive Principle:
Principle #3Local quality

2Loss of information

If diagonal scan order is used for encoding transform coefficients, then compression efficiency is improved, but encoding and decoding time increases due to pipelining problems

Engineering Contradiction:
Improvecompression efficiencyVSAvoidencoding time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The transform coefficient block is divided into multiple sub-blocks, and different scan orders are applied to different sub-blocks. Specifically, diagonal scan order is used for some sub-blocks to maintain compression efficiency, while horizontal or vertical scan orders are used for other sub-blocks to improve throughput and eliminate pipelining bottlenecks at corner positions.

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If context templates are designed to include neighboring positions for adaptive coding, then coding accuracy is improved, but pipelining capability deteriorates due to data dependency

Engineering Contradiction:
Improvecoding accuracyVSAvoidpipelining capability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The context modeling is segmented by applying different context templates to different sub-blocks. For sub-blocks where pipelining is critical (such as those at corner positions), simplified context templates are used that do not require future neighboring data. For other sub-blocks, full adaptive context templates are used to maximize coding accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Context templates are designed to use only previously decoded neighboring positions rather than future positions. This preliminary action approach ensures that all context data is available before encoding the current coefficient, enabling pipelining without data dependency bottlenecks.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2637405B1Adaptive bypass mode for coding absolute values of transform coefficients
Publication Date: 2018.06.13 BLACKBERRY LTD
  • EP2637405B1 patent drawingFigure 1
  • EP2637405B1 patent drawingFigure 2
  • EP2637405B1 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, using adaptive-threshold-based level coding. Threshold is set based upon level information from one or more previously-reconstructed coefficient groups in the transform unit. Threshold may be maximum number of level flags to decode for the coefficient group. Level information may include number of level flags decoded in previous coefficient groups. Previously-reconstructed coefficient groups may include coefficient group to the right and below the current coefficient group.