Entropy Slice Coding for Low-Delay Wavefront Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increased processing requirements of the HEVC standard and rising video resolution necessitate efficient methods for utilizing multi-core architectures in video coding, but existing parallelization techniques suffer from coding efficiency losses and high bitstream burdens, particularly in low-delay scenarios.
Innovation Solution
The proposed solution involves adapting probability estimations in entropy coding by incorporating statistics from spatially neighboring entropy slices, allowing for closer alignment with actual symbol statistics and reducing coding efficiency penalties, while enabling low-delay processing and wavefront processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If wavefront processing is used for parallelization, then processing speed is improved, but coding efficiency deteriorates due to interrupted spatial dependencies
Solution Approach 1:
The picture is divided into multiple entropy slices that can be processed in parallel. Each slice is further divided into parts that can be independently decoded, enabling wavefront processing while maintaining manageable dependency scopes within each slice.
Solution Approach 2:
Probability estimations are initialized at the beginning of each entropy slice using pre-defined values that are adapted to representative blend of sample array material. This preliminary initialization allows parallel processing to start immediately without waiting for full adaptation from previous slices.
Solution Approach 3:
Probability estimations are continuously adapted during the course of coding within each entropy slice based on actual symbol statistics. The adaptation feedback mechanism allows the system to converge toward optimal probability values even when starting from pre-defined initialization values.
2Loss of time
If probability estimations are initialized to pre-defined values, then processing delay is reduced, but coding efficiency deteriorates due to deviation from actual symbol statistics
Solution Approach 1:
Probability estimations are initialized at the beginning of each entropy slice using pre-defined values that are adapted to representative blend of sample array material. This preliminary initialization allows parallel processing to start immediately without waiting for full adaptation from previous slices.
Solution Approach 2:
Probability estimations are continuously adapted during the course of coding within each entropy slice based on actual symbol statistics. The adaptation feedback mechanism allows the system to converge toward optimal probability values even when starting from pre-defined initialization values.
3Device complexity
If probability adaptation is performed only within current entropy slice, then parallelization is simplified, but adaptation rate to actual symbol statistics decreases
Solution Approach 1:
The probability adaptation process is segmented into intra-slice adaptation (maintaining simplicity for parallelization) and inter-slice adaptation (improving convergence rate). Each slice maintains its own adaptation state while benefiting from initialization based on representative values from other slices.
Solution Approach 2:
Pre-defined probability initialization values act as intermediaries that carry information about typical symbol statistics across slices. These intermediary values provide a head start for probability adaptation without creating complex dependencies between slices.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The entropy coding of a current part of a predetermined entropy slice is based on, not only, the respective probability estimations of the predetermined entropy slice as adapted using the previously coded part of the predetermined entropy slice, but also probability estimations as used in the entropy coding of a spatially neighboring, in entropy slice order preceding entropy slice at a neighboring part thereof. Thereby, the probability estimations used in entropy coding are adapted to the actual symbol statistics more closely, thereby lowering the coding efficiency decrease normally caused by lower-delay concepts. Temporal interrelationships are exploited additionally or alternatively.