Entropy Slice Context Adaptation for Low-Delay Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in parallelization due to spatial dependencies between LCUs, leading to coding efficiency losses and increased bitstream burdens, particularly with the HEVC standard and increasing video resolutions, which hinder low-delay processing.
Innovation Solution
The proposed solution involves a decoder and encoder configuration that entropy decodes and encodes sample arrays by adapting probability estimations sequentially along entropy coding paths, utilizing previously decoded or encoded parts to improve coding efficiency and reduce delay, while allowing for parallelization without significant penalties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If spatial dependencies between LCUs are maintained for accurate probability adaptation, then coding efficiency is improved, but parallelization capability deteriorates
Solution Approach 1:
The image is divided into multiple entropy slices that can be processed in parallel. Each slice contains a subset of LCUs and can be decoded independently to a certain extent, allowing parallel processing while maintaining necessary spatial dependencies within each slice for accurate probability adaptation.
Solution Approach 2:
Different probability adaptation strategies are applied to different regions. Within each entropy slice, full spatial dependencies are maintained for accurate CABAC probability adaptation. Between slices, selective dependency handling enables parallelization. This local differentiation resolves the contradiction by allowing full dependencies where needed and reduced dependencies where parallelization is prioritized.
2Loss of energy
If full spatial dependencies are maintained across all LCUs, then coding efficiency is maximized, but processing delay increases
Solution Approach 1:
By segmenting the image into entropy slices, the processing can be organized to maintain full spatial dependencies within each slice (for coding efficiency) while enabling overlapping processing across slices (reducing delay). Multiple slices can be processed in parallel or with overlapping operations.
Solution Approach 2:
Probability models are initialized and prepared in advance for each entropy slice based on available information from previously decoded slices or reference frames. This preliminary preparation allows decoding to start earlier without waiting for complete spatial dependency resolution across the entire image, reducing processing delay while maintaining coding efficiency within each slice.
3Productivity
If entropy slices are processed in parallel, then productivity is improved, but coding efficiency deteriorates due to probability estimation inaccuracies
Solution Approach 1:
Each entropy slice processes spatial dependencies locally with high accuracy using adapted probability models, while parallel processing occurs between slices. The CABAC probability adaptation is performed accurately within each slice's local context, minimizing coding efficiency loss despite parallel processing.
Solution Approach 2:
Probability models are continuously adapted based on feedback from decoded data within each entropy slice. This feedback mechanism ensures accurate probability estimations are maintained even when slices are processed in parallel, preventing coding efficiency deterioration.
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.