Entropy Slice Probability Adaptation for Low-Delay Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving low-delay processing while maintaining coding efficiency due to the need to interrupt spatial dependencies between adjacent coding units, leading to inefficiencies in parallelization.
Innovation Solution
The proposed solution involves entropy decoding and encoding sample arrays by adapting probability estimations based on previously decoded neighboring slices, allowing for sequential initialization and continuous adaptation of probability estimations across slices, enabling low-delay processing with reduced coding efficiency penalties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If spatial dependencies between adjacent LCUs are maintained for accurate probability estimation, then coding efficiency is improved, but parallelization capability deteriorates
Solution Approach 1:
The video sequence is divided into multiple independent processing units (slices or wavefront rows) that can be decoded in parallel. Each unit contains a subset of LCUs that are processed independently, breaking the global spatial dependencies into localized segments. This segmentation allows multiple cores to work simultaneously on different segments while maintaining necessary local dependencies within each segment.
Solution Approach 2:
The patent introduces a new processing dimension by organizing LCUs into wavefront rows or diagonal processing orders instead of traditional raster scan. This dimensional reorganization allows dependencies to be satisfied within each wavefront row while enabling parallel processing across different rows, effectively transforming the dependency structure from a single-dimensional sequential chain to a multi-dimensional parallel structure.
2Loss of information
If probability estimations are adapted continuously along the coding path, then coding efficiency is improved, but processing delay increases
Solution Approach 1:
Probability estimations are initialized in advance using data from previously decoded frames or reference pictures before actual decoding begins. This preliminary initialization provides accurate starting probability values without requiring sequential adaptation during the decoding process, thereby reducing processing delay while maintaining coding efficiency.
Solution Approach 2:
The patent maintains continuous probability adaptation within each parallel processing unit (slice or wavefront row) while enabling discontinuous parallel execution across different units. The adaptation process continues uninterrupted within local segments, preserving coding efficiency, while the overall processing achieves low delay through parallel execution of multiple segments.
3Productivity
If wavefront processing is used to enable parallelization, then productivity is improved, but coding efficiency loss increases
Solution Approach 1:
The patent applies different probability estimation strategies to different local regions. Within each wavefront row or slice, continuous probability adaptation is maintained to preserve local coding efficiency. At the same time, independence between different wavefront rows or slices is enforced to enable parallel processing. This local differentiation allows the system to optimize for both parallelization and coding efficiency in their respective domains.
4Loss of information
If sequential entropy decoding of slices is performed, then parallelization is reduced, but coding efficiency is maintained
Solution Approach 1:
The patent implements dynamic probability estimation that adapts to the local characteristics of each slice or wavefront row. Probability models are updated dynamically within each parallel processing unit based on local statistics, allowing accurate coding efficiency to be maintained even when slices are processed in parallel rather than sequentially. This dynamic adaptation eliminates the need for strict sequential processing.
Data Source
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.