Slice-Type Context Resetting for Parallel Video Entropy Decoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

State-of-the-art video-coding methods like H.264/AVC and TMuC provide higher coding efficiency but at the expense of increased complexity, which can be mitigated by supporting parallel decoding, especially with advances in multi-core processors, but entropy decoding remains a bottleneck due to serial processing within slices.

Innovation Solution

The invention introduces the concept of entropy slices, allowing parallel decoding of video frames by partitioning reconstruction slices into self-contained entropy slices, where context models are reset at the start of each slice, enabling independent decoding and reducing dependency on neighboring slices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If video coding methods use macroblock prediction followed by residual coding to reduce temporal and spatial redundancy, then coding efficiency is improved, but decoder complexity increases

Engineering Contradiction:
Improvecoding efficiencyVSAvoiddecoder complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The picture is divided into multiple slices, and each slice is further divided into entropy slices. This segmentation allows the decoder to process smaller independent units in parallel, reducing the complexity of the overall decoding process while maintaining coding efficiency through the preserved spatial and temporal prediction structures.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If video coding methods use variable-sized coding tree blocks to provide flexibility, then adaptability is improved, but device complexity increases

Engineering Contradiction:
Improvecoding flexibilityVSAvoidencoder and decoder complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The variable-sized coding tree blocks are organized into fixed-structure entropy slices with predetermined maximum sizes. This segmentation approach allows the decoder to handle flexible block sizes within a standardized processing framework, maintaining adaptability while reducing complexity through consistent slice boundary handling and parallel processing capabilities.

Inventive Principle:
Principle #1Segmentation

3Reliability

If entropy decoding is performed serially within slices, then decoding accuracy is maintained, but decoding speed decreases

Engineering Contradiction:
Improvedecoding accuracyVSAvoiddecoding speed
Core Design Contradiction:
ReliabilityVSSpeed

Solution Approach 1:

Slices are divided into multiple entropy slices that are self-contained and can be decoded independently in parallel. Each entropy slice maintains the necessary context information for accurate decoding, allowing multiple decoding operations to proceed simultaneously without compromising decoding accuracy, thereby significantly improving overall decoding speed.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Context models are reset at the start of each entropy slice, preparing the decoding state in advance for independent parallel processing. This preliminary initialization ensures that each entropy slice can be decoded accurately without requiring sequential processing of previous slices, enabling parallel execution while maintaining reliability.

Inventive Principle:
Principle #10Preliminary action

4Speed

If context models are reset at the start of each entropy slice to enable parallel decoding, then decoding speed is improved, but loss of information may occur

Engineering Contradiction:
Improvedecoding speedVSAvoidcontext information
Core Design Contradiction:
SpeedVSLoss of information

Solution Approach 1:

The picture is segmented into entropy slices with carefully designed boundaries that ensure all necessary context information is contained within each slice or can be independently reconstructed. This segmentation strategy enables parallel decoding with context model resets while preventing information loss by ensuring each entropy slice is self-contained with respect to the entropy coding process.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4380157B1Context initialization based on slice type
Publication Date: 2025.09.10 VELOS MEDIA INT LTD
  • EP4380157B1 patent drawingFigure 1
  • EP4380157B1 patent drawingFigure 2
  • EP4380157B1 patent drawingFigure 3

AI summary

A method for decoding a video frame of a video sequence comprising: (a) receiving a slice header; (b) identifying a slice type; (c) receiving a first flag indicating a initialization method used to initialize a context; and (d) initializing the context associated with the slice type using one of a first initialization method, second initialization method and third initialization method.