Video Coding Data Segmentation for Encoder-Decoder Synchronization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like JEM have rudimentary methods for predicting ALF parameter values and CABAC probability states across pictures, leading to issues such as encoder-decoder mismatches, errors, and inefficiencies in handling temporal layer pruning and random access operations.
Innovation Solution
The proposed solution involves managing data predicted across pictures by separating it into sample data and non-sample data types, allowing for selective storage and removal of data in decoder and encoder memories based on bitstream information, thereby ensuring accurate and efficient decoding and encoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If ALF parameter values are predicted from reference pictures using FIFO queue, then bitrate is reduced, but encoder-decoder mismatches and errors occur during temporal layer pruning and random access operations
Solution Approach 1:
The patent segments the reference picture data into two distinct types: sample data (pixel values) and non-sample data (ALF parameters, CABAC states). This segmentation allows different retention policies to be applied to each type, resolving the contradiction by enabling bitrate reduction through selective non-sample data removal while preserving sample data for accurate reconstruction and matching.
Solution Approach 2:
The patent applies different quality levels and retention strategies to different data types locally. Sample data is fully retained to maintain reconstruction accuracy, while non-sample data is selectively removed based on picture type and usage. This local differentiation resolves the contradiction by optimizing bitrate without compromising encoder-decoder synchronization.
2Measurement precision
If sample data is stored in decoder memory for all previously decoded pictures, then accurate prediction is enabled, but memory consumption increases
Solution Approach 1:
The patent extracts and removes non-sample data (ALF parameters, CABAC states) from the decoder memory when the corresponding sample data is removed. This extraction principle reduces memory consumption by eliminating redundant non-sample data while preserving sample data necessary for accurate prediction and reconstruction.
Solution Approach 2:
By segmenting reference picture data into sample and non-sample components, the patent enables selective retention of only the essential sample data in memory, while non-sample data is discarded when no longer needed. This segmentation resolves the memory consumption issue while maintaining prediction accuracy.
3Productivity
If non-sample data is retained in memory after sample data removal, then cross-picture prediction efficiency is improved, but encoder-decoder mismatches occur
Solution Approach 1:
The patent extracts and removes non-sample data from memory simultaneously with sample data removal. This coordinated extraction ensures that both encoder and decoder maintain consistent memory states, preventing mismatches while allowing efficient memory management and maintaining decoding productivity.
Solution Approach 2:
The patent implements feedback mechanisms where the encoder and decoder synchronize their memory management decisions based on bitstream information. This feedback ensures that non-sample data is removed at the same points in both encoder and decoder, maintaining reliability while preserving decoding efficiency.
Data Source
AI summary
Data that is predicted across pictures in a video sequence is managed by separating the data into multiple data types. Instead of keeping all data associated with a decoded picture, such as picture sample values and motion vector data, data associated with a decoded picture is split by data type to enable storing only a subset of all data associated with a decoded picture.