VCM Bitstream Type Signaling for Machine Vision Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies are not optimized for machine-oriented tasks in artificial intelligence services, leading to inefficiencies in encoding and decoding processes.
Innovation Solution
A method and apparatus for encoding and decoding VCM bitstreams that include determining optimization methods based on type information and encoding/decoding efficiency, with a focus on machine-oriented image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing image compression technologies are used for machine-oriented tasks, then high-resolution and high-quality image processing is achieved, but encoding and decoding efficiency for artificial intelligence services deteriorates
Solution Approach 1:
The patent changes the optimization parameters of image compression technology from human-vision-oriented parameters to machine-task-oriented parameters. By introducing type information that indicates optimization methods and adjusting encoding/decoding parameters based on machine learning task requirements, the system achieves both high image quality and improved encoding efficiency for AI services
Solution Approach 2:
The patent makes the compression system dynamic by allowing selection of different optimization methods based on task type. The type information enables the system to adaptively switch between different encoding/decoding configurations, making the system flexible enough to handle various machine learning tasks while maintaining efficiency
2Manufacturing precision
If existing image compression technologies are used for machine-oriented tasks, then high-resolution and high-quality image processing is achieved, but decoding efficiency for artificial intelligence services deteriorates
Solution Approach 1:
The patent optimizes decoding parameters specifically for machine learning tasks by introducing type information that indicates the intended use. This allows the decoding process to be tailored to machine-oriented requirements, reducing decoding time while maintaining the quality needed for AI processing
Solution Approach 2:
The patent performs preliminary optimization by embedding type information during encoding that guides the decoding process. This preliminary action allows the decoder to pre-configure optimal processing paths for machine learning tasks, reducing the time required during actual decoding operations
3Productivity
If optimization methods are added to improve machine-oriented efficiency, then encoding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent creates a universal optimization framework where type information serves multiple functions: it indicates the intended machine learning task, guides encoding parameter selection, and directs decoding optimization. This multi-functionality reduces the need for separate complex systems for different tasks, thereby limiting the increase in device complexity while improving encoding efficiency
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are an encoding/decoding method, a device, and a computer-readable recording medium storing bitstreams generated by the encoding method. The decoding method according to the present disclosure may be a decoding method performed by a decoding device and comprising the steps of: acquiring, from a bitstream, type information indicating at least one of optimization methods applicable to encoding; and determining, on the basis of the type information, the at least one optimization method applied to the encoding.