Visual Residual Coding With Multi-Threaded Hyperprior Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression technologies, both classical and neural-network-based, face challenges in achieving optimal coding efficiency and effectiveness, particularly in modeling probability distributions and handling spatial dependencies in latent variables, which affects the compression ratio and reconstruction quality.
Innovation Solution
A method and apparatus for visual data processing that utilize a plurality of threads to code residual information, incorporating a hyperprior model and autoregressive components to improve coding efficiency by modeling spatial dependencies and probability distributions in visual data, and a decoupled framework for independent decoding of residual information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single thread is used for coding residual information, then the device complexity is low, but the coding efficiency and productivity are insufficient
Solution Approach 1:
The patent divides the coding task into multiple independent threads, each responsible for coding specific residual information blocks. This segmentation enables parallel processing of different regions or types of residual data, thereby improving coding efficiency without requiring complex inter-thread coordination for each coding operation.
Solution Approach 2:
The patent performs preliminary organization and allocation of residual information to different threads before the actual coding process begins. This pre-arrangement includes determining thread assignments, preparing data structures, and setting up coding parameters in advance, which reduces runtime complexity and enables efficient parallel execution.
2Productivity
If multiple threads are used for coding residual information, then the coding efficiency is improved, but the device complexity increases
Solution Approach 1:
The patent introduces an intermediary control mechanism that manages multiple coding threads through a unified interface. This intermediary layer handles thread coordination, resource allocation, and result aggregation, allowing complex multi-threaded operations to be controlled through simplified procedures and reducing overall system complexity.
3Loss of information
If spatial dependencies are not modeled, then the coding process is simple, but the compression ratio and reconstruction quality deteriorate
Solution Approach 1:
The patent implements feedback mechanisms where previously coded residual information and reconstructed visual data are used to inform subsequent coding decisions. This feedback loop enables the system to adapt to spatial dependencies and statistical characteristics of the visual data, improving compression efficiency and reconstruction quality without requiring overly complex explicit modeling.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for visual data processing. In the method, for a conversion between a current visual unit of visual data and a bitstream of the visual data, a plurality of threads for coding residual information of the current visual unit is determined. The conversion is performed based on the plurality of threads.